Proportionate Sign Subband Adpative Filtering Algorithm for Network Echo Cancellers

Size: px
Start display at page:

Download "Proportionate Sign Subband Adpative Filtering Algorithm for Network Echo Cancellers"

Transcription

1 Journal of Advances n Computer Networks, Vol. 3, No. 2, June 2015 Proportonate Sgn Subband Adpatve Flterng Algorthm for Network Echo Cancellers J-Hye Seo, Sang Mok Jung, and Poo Gyeon Park are chosen as standard for the NEC. Wth long mpulse response, however, most of the coeffcents are close to zero; t means the typcal network mpulse response s sparse. Second, the exctaton sgnals used for NEC are speech sgnals whch are hghly correlated and non-gaussan process, so many adaptve flterng algorthms suffer from reduced convergence rate. In addton, the background nose can also be hghly non-statonary or strong nterferences. he thrd aspect to be consdered n NEC s double-talk scenaro whch means that far-end and near-end speech are actve smultaneously. In ths case, a near-end speech acts lke a large dsturbance of the flter, so t may causes extremely slow convergence or dvergence. When developng adaptve flterng algorthms used for NEC, these challengng problems should be consdered. An deal algorthm not only should have fast convergence rate and good trackng performance but also acheves small steady-state estmaton error even for long-length echo path. hese problems should be mantaned n the case of non-statonary exctaton sgnal. Furthermore, the algorthm should be robust aganst varaton of the mcrophone sgnal (background nose varatons and double-talk stuaton). he conventonal normalzed least mean squares (NLMS) algorthm s the most wdely used adaptve flterng algorthm due to ts smplcty and robustness [1], [2]. However, for NEC, the NLMS algorthm suffers from reduced convergence rate due to long length of the flter. It s possble to consder the recursve least squares (RLS) algorthm and the affne projecton algorthm (APA) [3] for obtanng fast convergence rate, but they are not approprate for NEC applcaton n practce because of ther heavy computatonal complexty. hus, there are requrements for adaptve flterng algorthms wth fast convergence rate and low computatonal complexty to replace the NLMS algorthm. he network echo path mpulse response s known as sparse n nature. hat s, only a small porton of the coeffcents s actve and other coeffcents are zero or unnotceably small values (nactve). Many adaptve flterng algorthm take advantage of sparseness the echo path and propose some methods to mprove the convergence rate and robustness or lower computatonal complexty. he proportonate NLMS (PNLMS) [4] has been developed to acheve fast ntal convergence rate. It updates coeffcents of the weght vector by assgnng dfferent step szes to dfferent coeffcents based on ther magntude of current estmates. For a large coeffcent, a large value of step sze s assgned n the update process, so ths method allows to obtan fast ntal convergence rate for sparse mpulse responses relatvely ndependent of ther length for sparse mpulse responses. However, the PNLMS algorthm slows down after ts fast Abstract hs paper proposes a proportonate sgn subband adaptve flterng (PSSAF) and an mproved PSSAF (IPSSAF) algorthm for network echo cancellers that deal wth mpulsve nterferences and sparse echo paths. Based on a sgn subband adaptve flterng (SSAF) algorthm that s robust aganst mpulsve nterferences, the proposed algorthm mnmzes L1-norm of the subband a posteror error vector subject to a weghted constrant on the flter coeffcents. A postve defnte weghtng matrx s used for the constrant. In ths paper, we adopt a dagonal proportonate matrx for the constrant to acheve fast ntal convergence rate when the mpulse response s sparse. he components of t s roughly proportonal to the absolute value of current estmate of the flter, so the resultng algorthm s called PSSAF algorthm. o acheve fast ntal convergence rate even for rather non-sparse mpulse responses, an mproved proportonate matrx s also used for the constrant, and the resultng algorthm s named as IPSSAF algorthm. Expermental results show that the proposed algorthms are more robust aganst hghly correlated nput sgnals, mpulsve nterferences, and double-talk than the orgnal normalzed subband adaptve flterng algorthm and the class of proportonate subband adaptve flterng algorthms. Index erms Adaptve flters, proportonate, sgn algorthm, subband adaptve flterng algorthm, weghted constrant, weghtng matrx. I. INROUCION A network echo cancellers (NEC) s an adaptve flter that s one of the best solutons to control acoustc echos of the voce communcaton networks. An adaptve flter estmates the echo path of the network and generates a replca of the room mpulse response, and then the output of the flter s subtracted from the near-end sgnal to obtan clean sgnals. he scheme of the NEC s the same as a classcal system dentfcaton, but there are several challengng problems for modern voce transmsson over telephone networks. he frst one s about the characterstcs of the echo path. he length of the echo path for NEC s extremely long due to long delays caused by large scale network. hus, the adaptve flter usually requres large number of taps, and 512 or 1024 taps Manuscrpt receved September 2, 2014; revsed March 23, hs research was supported by the MSIP (Mnstry of Scence, IC and Future Plannng), Korea, under the I Conslence Creatve Program (NIPA-2014-H ) supervsed by the NIPA (Natonal I Industry Promoton Agency). J-Hye Seo s wth the vson of I Convergence Engneerng, Pohang Unversty of Scence and echnology, Pohang, Gyeongbuk, Korea (e-mal: bmclubhs@postech.ac.kr). Sang Mok Jung and Poo Gyeon Park are wth the epartment of Electrcal Engneerng, Pohang Unversty of Scence and echnology, Pohang, Gyeongbuk, Korea (e-mal: llus2@postech.ac.kr, ppg@postech.ac.kr). OI: /JACN.2015.V

2 Journal of Advances n Computer Networks, Vol. 3, No. 2, June 2015 ntal convergence, and t has slower convergence rate than NLMS when the network mpulse response s not sparse enough. hs problem has been addressed n several papers, and modfcatons to PNLMS algorthm s studed to mprove the performance. he mproved PNLMS (IPNLMS) [5] s proposed to avod degradaton of the performance for the non-sparse mpulse response. It gves more optmal step szes for gven sparseness by usng a control factor, so mantans ts fast convergence rate even n the case of non-sparse mpulse responses. o keep the fast convergence rate durng the whold adaptaton process, the µ-law PNLMS (MPNLMS) s proposed n [6]. Also, the mproved IPNLMS (IIPNLMS) [7] s proposed where the step szes assgned for dfferent coeffcents are determned by two dfferent relatonshps dependng on whether the coeffcent s consdered as actve or nactve. hese concept of proportonate-type algorthms are extended to a normalzed subband adaptve flterng (NSAF) algorthm [8] n [9]. he NSAF algorthm decomposes the nput sgnals nto multple subbands n order to whten the nput sgnals n each subband, so the algorthm s sutable to acheve faster convergence rate than the NLMS algorthm. he computatonal complexty of the NSAF algorthm s not heavy comparng to the RLS or APA, and thus, the proportonate-type subband adaptve flterng algorthms are proper to use n NEC applcatons. In addton to fast convergence rate and low complexty, a robustness can be an consderable ssue for adaptve flterng algorthms. In partcular, double-talk s a bg challengng problem for NEC. o deal wth double-talk stuaton, a double-talk detector () can be used typcally. However, t s not desrable to rely only on because the NEC has tght requrements for detectng double-talk. hus, several adaptve flterng algorthms are developed focusng on the robustness to double-talk [10]-[14]. One of the method s usng L 1 -norm mnmzaton of error because L 1 -norm algorthms are especally robust aganst mpulsve nose such as speech. he basc algorthm usng L 1 -norm s the normalzed sgn algorthm (NSA). he NSA provdes robustness aganst mpulsve nose, but the convergence rate of t becomes slower than the NLMS algorthm. Recently, a sgn subband adaptve flterng (SSAF) algorthm s suggested n [15] whch extends a concept of L 1 -norm error mnmzaton to the NSAF. he SSAF algorthm has good robustness comparng to NSAF algorthm, and acheves rather faster convergence rate than the NSA. In ths paper, we adopt the proportonate approach wth the SSAF to obtan faster convergence for sparse network mpulse responses. Based on the weghted constrant to the weght error vector, we obtaned a weght update equaton for the proposed algorthm mathematcal dervaton. wo types of proportonate matrx based on the PNLMS [4] and IPNLMS [5] are used as a weghtng matrx, and the resultng algorthms are called proportonate sgn subband adaptve flterng (PSSAF) algorthm and mproved PSSAF (IPSSAF) algorthm, respectvely. he PSSAF and IPSSAF algorthm show fast convergence rate for sparse mpulse response, and they also show the robustness aganst hghly correlated nput sgnal, nterferences, and double-talk. Especally, the IPSSAF algorthm also acheves fast convergence rate even for the non-sparse mpulse response. he computatonal complexty of the proposed algorthms are slghtly hgher than the SSAF algorthm, but lower than the NSAF and the famly or proportonate-type NSAF algorthms. Fg. 1. Structure of a network echo canceller (NEC). II. BACKGROUNS A. Network Echo Cancellers Consder a NEC scheme shown n Fg. 1. In Fg. 1, u( n ) denotes the far-end sgnal, and η ( n)= z( n) + v( n) s a near-end sgnal where z( n ) and v( n ) are the near-end speech and background nose sgnal, respectvely. he true echo path s denoted as w wth length M, and the estmaton of t can be o obtaned by mnmzng the dfference between the desred output d( n ) and the estmated flter output y( n ) such that e( n)= d( n) y( n). he desred output that contans the echo, the near-end sgnal can be represented as d( n) = u ( n) w +η( n), and the flter output, the replca of the o echo, can be expressed as y( n) = u ( n) w ( n), respectvely, where u( n) =[ u( n), u( n 1),, u( n M + 1)] s the far-end sgnal vector and w( n) =[ w ( n), w ( n),, w ( n)] s the 0 1 M 1 estmaton of the mpulse response at tme ndex n. B. Sgn Subband Adaptve Flterng Algorthm Fg. 2 shows the conventonal subband adaptve flterng structure. he sgnals u( n) and d( n ) are decomposed nto u ( n ) and d ( n ) by analyss flters H( z ) = 0,1,, N 1. he subband nput sgnals u ( n ) are fltered by the adaptve flter W( z ) and the flter outputs y ( n), = 0,1,, N 1 are obtaned. By decmatng d ( n ) and y ( n ), d ( k ) and, y ( k ) are generated, and t s easy to note that, y ( k) = u ( k) w ( k), where w ( k) s the tap weght vector of, W( z) and u ( k) =[ u ( kn), u ( kn 1),, u ( kn M + 1)]. We use n and k to ndex the orgnal and decmated sgnal, respectvely. efne the subband output error vector e ( ) and subband a posteror error vector e ( k ) as p k e ( k) = d ( k) U ( k) w ( k) (1) e ( k) = d ( k) U ( k) w ( k + 1) (2) p where 0, 1, N 1, d ( k) = [ d ( k), d ( k),, d ( k)] and U( k) = [ u ( k), u ( k),, u ( k)]. 0 1 N 1 100

3 Journal of Advances n Computer Networks, Vol. 3, No. 2, June 2015 Fg. 2. Structure of NSAF algorthm. he sgn subband adaptve flterng (SSAF) algorthm [15] can be obtaned by mnmzng the L 1 -norm of the a posteror error vector wth a constrant on the flter coeffcents, Mnmzng Subject to 2 2 w( k + 1) w ( k ) µ (6) Σ( k) Subject to d ( k ) U ( k ) w ( k + 1) 1 ( k 1) ( k ) 2 µ 2 2 w + w (3) Usng the method of Lagrange multplers, the cost functon can be obtaned as J( w( k + 1))= d ( k) U ( k) w ( k + 1) 1 + γ [ w( k + 1) w ( k) µ ] (7) 2 2 Σ( k) where µ s a postve step sze that does not change dramatcally. Usng Lagrange multplers to solve ths constraned optmzaton problem and replacng e ( k ) to p e ( ) k, the weght update equaton of the SSAF can be obtaned as follows: w( k + 1)= w( k) + µ U( k) sgn( e ( k)) (4) sgn( e ( k)) U ( k) U( k) sgn( e ( k)) Usng the dagonal assumpton, that s, U ( k ) U ( k) = dag[ u ( k) u ( k), u ( k) u ( k),, u ( k) u ], we can wrte (4) as N 1 N 1 w( k + 1)= w( k) + µ U( k) sgn( e ( k)) N 1 u =0 where δ s a regularzaton parameter. ( k) u ( k) + δ III. PROPORIONAE SIGN SUBBAN AAPIVE FILERING ALGORIHM Instead of usng L 2 -norm to the constrant on the flter coeffcents n (3), we can use a weghted norm wth a postve defnte weghtng matrx Σ ( k) as Mnmzng d ( k ) U ( k ) w ( k + 1) 1 (5) where γ s a Lagrange multpler. Settng the dervatve of (7) wth respect to the w ( k + 1) equal to zero, we get 1 1 w( k + 1) = w( k) + Σ ( k) U( k) sgn( e ( k)). (8) p 2γ Snce a posteror error vector e ( ) s not accessble before the update, we need to approxmate t to a pror error vector e ( k ). hen, (8) can be replaced wth p k 1 1 w( k + 1) = w( k) + Σ ( k) U( k) sgn( e ( k)). (9) 2γ Substtutng (9) to the constrant n (7), we obtan 1 = γ 1 2 ( sgn e ( k)) U ( k) Σ ( k) U( k) sgn( e ( k)) µ (10) Substtutng (10) nto (9), the update equaton of the flter coeffcents can be expressed as + µ w( k + 1) = w ( k) ( k) ( k) sgn( ( k)) 1 Σ U e sgn k k k k sgn k 1 ( e ( )) U ( ) Σ ( ) U( ) ( e ( )) (11) We can choose the matrx Σ ( k) dependng on the 1 envronment. hus, we choose Σ( k) = G ( k) = dag[1/ g ( k)] M l 1 l=0 as a weghtng matrx, and then (11) can 101

4 Journal of Advances n Computer Networks, Vol. 3, No. 2, June 2015 process and a Gaussan process,.e., z ( k ) = w( k ) n ( k ), be expressed wth the matrx G (k ) as where n ( k ) s a whte Gaussan random sequence wth zero w ( k + 1) = w ( k ) mean and varance σ n2, and w(k ) s a Bernoull process wth G ( k ) U( k ) sgn(e ( k )) +µ the probablty mass functon gven as P( w) = 1 Pr for (12) sgn(e ( k ))U ( k )G ( k ) U( k ) sgn(e ( k )) w = 0, and P ( w) = Pr for w=1. he average power of a BG process s Pr σn2. Keepng the average power constant, a he components of G ( k ) can be determned accordng to [4] and [5] as gl (k ) = γ l (k ) BG process s spker when Pr s smaller. It reduces to a Gaussan process when Pr = 1. he MS,.e., E w o w ( k ) 2, s an ndcator of the performance of the (13) M 1 algorthms, and t s obtaned by ensemble averagng over 20 ndependent trals. In addton, the regularzaton factors are set to δ = σ u2 for the NSAF, PSAF, SSAF, and PSSAF γ (k ) / M l l =0 algorthms, and δ IP = (1 α ) / 2δ where for the IPSAF and IPSSAF algorthms where σ s a varance of nput sgnal. 2 u γ l (k ) = max( ρ max(q, w0 (k ),, wm 1 (k ) ), wl (k ) ) (14) Fg. 4 shows the MS curves for our proposed algorthms and several other algorthms under consderaton for comparson. AR(1) process generated usng G1 was used as or gl (k ) = 1 α 2 + (1 + α ) wl (k ) (2 w ( k ) 1 +ε ) / M. an nput sgnal, and SNR was set to 30dB. Consderng the strong nterference wth Pr = 0.01, the proposed PSSAF and IPSSAF algorthms show faster convergence rate than the SSAF algorthm, whle the NSAF, PSAF, and IPSAF algorthm dverges because of ther L2-norm mnmzaton of the error vector. (15) he resultng algorthms dependng on the component g l ( k ) n two ways are called proportonate sgn subband adaptve flterng (PSSAF) and mproved PSSAF(IPSSAF) algorthms, repectvely. In (14), the parameter ρ prevents wl (k ) from stallng when t s much smaller than the largest coeffcent and q regularzes the updatng when all coeffcents are zero for the PSSAF algorthm. In (15), the parameter α controls the behavor of the IPSSAF algorthm. When α = -1, the IPSSAF and SSAF are dentcal, and the IPSSAF behaves lke the PSSAF when α gets closer to 1. IV. SIMULAION RESULS A. Performance Comparson In ths secton, we dd several smulatons to verfy the performance of our proposed algorthm. he echo paths used for smulatons have 512 coeffcents and two typcal mpulse responses are depcted n Fg. 3. wo nput sgnals are used n the smulatons: the frst one s obtaned by flterng a whte zero-mean Gaussan random sequence through the G1 ( z ) = 1 / (1 0.9 z 1 ), and the second one s a speech sgnal. Fg. 3. Room mpulse response of the echo path for NEC. (a) Sparse mpulse response. (b) spersve mpulse response. he measurement nose, v(k ), s added to c ( k ) wth a sgnal-to-nose rato (SNR) defned by E[c 2 (k )] 2 E[v (k )] SNR = 10 log10 where c( k ) = u ( k ) w o. he near-end sgnal, z ( k ), s a strong mpulsve nterference wth a sgnal-to-nterference rato (SIR) of -30dB and Bernoull-Gaussan (BG) dstrbuton s used for modelng the nterference sgnal. he BG dstrbuton s generated as the product of a Bernoull Fg. 4. MS curves for the NSAF, PSAF, IPSAF, SSAF, and proposed PSSAF and IPSSAF algorthm wth nterference for sparse mpulse response. Input sgnal s AR(1), and SNR = 30dB. 102

5 Journal of Advances n Computer Networks, Vol. 3, No. 2, June to make the IPSSAF algorthm perform robust to sparseness of the network mpulse responses. Fg. 5. MS curves for the SSAF, PSSAF and IPSSAF algorthm wth nterference for sparse mpulse response. Input sgnal s AR(1), and SNR = 30dB. Echo path s changed abruptly at k = B. rackng Another possble scenaro n NEC s the change of the echo path. he result for ths scenaro s depcted n Fg. 6, where the sparse echo path (Fg. 3(a)) was abruptly changed at k = to the dspersve mpulse response (Fg. 3(b)). Envronment for ths smulaton s the same as n Fg. 4, and the step szes for the SSAF, PSSAF, and IPSSAF were set to make them acheve the same steady-state MS. In Fg. 5, the IPSSAF algorthm acheves the fastest trackng performance among the comparng algorthms, and the PSSAF algorthm acheves better trackng performance than the SSAF algorthm, where the other algorthms track the changed echo path slowly. Fg. 6. MS curves for the PSSAF algorthm for dfferent ρ for µ =0.01. he smulaton envronment s the same as n Fg. 5. Impulse responses n Fg. 3 were used as the (a) sparse and (b) dspersve mpulse responses, respectvely. C. Parameters o verfy the effects of the proportonate matrx for sparse and dspersve echo paths, we dd addtonal smulatons by varyng parameters used for proportonate matrx. Fg. 6, Fg. 7 depct the results of the smulaton for the PSSAF and IPSSAF algorthms, respectvely. AR(1) nput and BG nterference was used n the smulatons for mpulse responses shown n Fg. 3, and the step sze was set to µ = In Fg. 6(a), the PSSAF algorthm wth small ρ(ρ = 0.01) leads to fast convergence rate for sparse mpulse response because more proportonalty for the flter PSSAF algorthm wth larger value of ρ acheves better performance for dspersve mpulse response because larger ρ makes lttle proportonalty, and the algorthm updates the flter coeffcents wth a smlar rate. We verfed that the value of ρ consderably affects the performance of the PSSAF algorthm dependng on the sparseness of the network mpulse responses. We also dd experment about the effect of α for the IPSSAF algorthm n Fg. 7. In Fg. 7(a), as the value of α get closer to -1, the convergence rate of the IPSSAF algorthm get slower for sparse mpulse response. he convergence behavor of the IPSSAF for dspersve mpulse response s shown n Fg. 7(b), and we can notce that the algorthm shows smlar ntal convergence rate regardless of the value of α. As the value of α ncreases, the convergence rate of the IPSSAF algorthm get slower after ntal convergence, but the steady-state estmaton error of the algorthm wth dfferent α s not much dfferent for dspersve mpulse response. he effect of α s less senstve than the effect ρ of the PSSAF algorthm, and t s good to choose α = 0 or α = Fg. 7. MS curves for the IPSSAF algorthm for dfferent α for µ =0.01. he smulaton envronment s the same as n Fg. 5. Impulse responses n Fg. 3 were used as the (a) sparse and (b) dspersve mpulse responses, respectvely.. ouble-alk Scenaro In ths subsecton, we show the performance of the proposed and competng algorthms n the double-talk stuaton. Expermental envronment s the same as Fg. 4 except exctaton sgnal, and mpulse response n Fg. 3(a) s used as the network echo path. Speech sgnals used for double-talk scenaro are depcted n Fg. 8; far-end and near-end speech, respectvely. ouble-talk happen n the perod wth sample [2.0, 4.0] 104. As shown n Fg. 9 and Fg. 103

6 Journal of Advances n Computer Networks, Vol. 3, No. 2, June 2015 (IPSSAF) algorthm. he resultng PSSAF and IPSSAF algorthm showed faster convergence rate than the conventonal L1 algorthm for sparse network mpulse response. Moreover, they acheved more robust performance aganst mpulsve nterferences and double-talk than the exstng algorthms. 10, the NSAF, PSAF, and IPSAF show non-robust performance aganst double-talk dsturbance wthout and wth nterferences. In contrast, the NSA, SSAF, PSSAF and IPSSAF algorthm, a famly of sgn algorthms, show somewhat robust performance aganst both mpulsve nterferences and double-talk dsturbance. Among these famly of sgn algorthms, the proposed PSSAF and IPSSAF acheves faster convergence rate and smaller steady-state estmaton error than the NSA and SSAF algorthm. Wthout any double-talk detector, the proposed PSSAF and IPSSAF show better robustness to double-talk. ACKNOWLEGMEN hs research was supported by the MSIP (Mnstry of Scence, IC and Future Plannng), Korea, under the I Conslence Creatve Program (NIPA-2014-H ) supervsed by the NIPA (Natonal I Industry Promoton Agency). REFERENCES [1] [2] [3] [4] Fg. 8. Speech sgnals used n the double-talk scenaro. (a) Far-end speech (b) Near-end speech. [5] [6] [7] [8] [9] [10] Fg. 9. MS curves for the NSAF, PSAF, IPSAF, NSA, SSAF, and proposed PSSAF and IPSSAF algorthm for double-talk scenaro wthout mpulsve nterferences. [11] [12] [13] [14] Fg. 10. MS curves for the NSAF, PSAF, IPSAF, NSA, SSAF, and proposed PSSAF and IPSSAF algorthm for double-talk scenaro wth mpulsve nterferences. [15] S. Haykn, Adaptve Flter heory, 1st ed. Upper Saddle Rver, NJ: Prentce-Hall, A. H. Sayed, Fundamentals of Adaptve Flterng, New York: Wley, K. Ozek and. Umeda, An adaptve flterng algorthm usng an orthogonal projecton to an affne subspace and ts propertes, Electroncs and Communcatons n Japan, vol. 67-A, no. 5, pp , L. uttweler, Proportonate normalzed least-mean-squares adaptaton n echo cancelers, IEEE rans. on Speech and Audo Processng, vol. 8, no. 5, pp , September J. Benesty and S. L. Gay, An mproved PNLMS algorthm, n Proc. IEEE Internatonal Conference on Acoustcs, Speech, and Sgnal Process, vol. 2, pp , May H. eng and M. oroslovack, Proportonate adaptve algorthms for network echo cancellaton, IEEE rans. on Sgnal Processng, vol. 54, pp , May P. A. Naylor, J. Cu, and M. Brookes, Adaptve algorthms for sparse echo cancellaton, Sgnal Processng, vol. 86, pp , K. A. Lee, W. S. Gan, and S. M. Kuo, Subband Adaptve Flterng: heory and Implementaton, Wley, M. S. E. Abad, Proportonate normalzed subband adaptve flter algorthms for sparse system dentfcaton, Sgnal Processng, vol. 87, no. 7, pp , July J. Benesty, H. Rey, L. R. Vega, and S. ressens, A nonparametrc VSS NLMS algorthm, IEEE Sgnal Processng Letters, vol. 13, no. 10, pp , October Gansler, S. L. Gay, M. M. Sondh, and J. Benesty, ouble-talk robust fast convergng algorthms for network echo cancellaton, IEEE rans. on Speech and Audo Processng, vol. 8, no. 6, pp , November Waterschoot, G. Rombouts, P. Verhoeve, and M. Moonen, ouble-talk-robust predcton error dentfcaton algorthms for acoustc echo cancellaton, IEEE rans. on Sgnal Processng, vol. 55, no. 3, pp , March C. Paleologu, J. Benesty, and S. Cochna, A varable step-sze affne projecton algorthm desgned for acoustc echo cancellaton, IEEE rans. on Audo, Speech, and Language Processng, vol. 16, no. 8, pp , November Z. Yang et al., Proportonate affne projecton sgn algorthms for network echo cancellaton, IEEE rans. on Audo, Speech, and Language Processng, vol. 19, no. 8, pp , November J. N and F. L, Varable regularsaton parameter sgn subband adaptve flter, Electroncs Letters, vol. 46, no. 24, pp , November J-Hye Seo receved the B.Sc. degree n electrcal engneerng from Pohang Unversty of Scence and echnology (POSECH), Pohang, Korea, n 2010, and the M.Sc. degree from the vson of I Convergence Engneerng from POSECH. Currently, she s a Ph.. canddate n the vson of I Convergence Engneerng, POSECH, n Her man research nterests nclude sgnal processng, adaptve flterng algorthm, and mage V. CONCLUSION In ths paper, we proposed proportonate-type sgn subband adaptve flterng algorthms for network echo cancellaton. he proposed algorthm mnmzes L1-norm of the subband a posteror error vector subject to a weghted constrant. Applyng proper postve-defnte weghtng matrx to the weghted constrant, we obtaned proportonate sng subband adaptve flterng (PSSAF) algorthm and mproved PSSAF processng. 104

7 Journal of Advances n Computer Networks, Vol. 3, No. 2, June 2015 mplementatons. Sang Mok Jung receved the B.Sc. and M.Sc. degrees n electrcal engneerng from Pohang Unversty of Scence and echnology (POSECH), Pohang, Korea, n 2010 and 2012, respectvely. Currently, he s a Ph.. degree canddate n the vson of Electrcal Engneerng, POSECH. Hs man research nterests nclude sgnal processng, adaptve flterng algorthm, mage nspectons, network programmng, and ther Poo Gyeon Park receved the B.S. degree and M.S. degree n control and nstrumentaton engneerng from Seoul Natonal Unversty, Korea, n 1988 and 1990, respectvely, and the Ph.. degree n electrcal engneerng from Stanford Unversty, Stanford, CA, n Snce 1996, he has been afflated wth the vson of Electrcal Engneerng, Pohang Unversty of Scence and echnology, where he s currently a professor. Hs current research nterests nclude robust, LPV, and network-related control theores, delayed systems, fuzzy systems, sgnal processng, and wreless communcatons for personal area network (PAN). 105

Pop-Click Noise Detection Using Inter-Frame Correlation for Improved Portable Auditory Sensing

Pop-Click Noise Detection Using Inter-Frame Correlation for Improved Portable Auditory Sensing Advanced Scence and Technology Letters, pp.164-168 http://dx.do.org/10.14257/astl.2013 Pop-Clc Nose Detecton Usng Inter-Frame Correlaton for Improved Portable Audtory Sensng Dong Yun Lee, Kwang Myung Jeon,

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

adaptive filter. However, the transient behavior of the INSAF algorithm has not been studied. In this paper, we focus on studying the

adaptive filter. However, the transient behavior of the INSAF algorithm has not been studied. In this paper, we focus on studying the Performance Analyses of Nose-Robust Normalzed Subband Adaptve Flter Algorthm 1 Y Yu a, Haquan Zhao b, Badong Chen c, Wenyuan Wang b, Lu Lu d Abstract: hs paper studes the statstcal models of the nose-robust

More information

CHAPTER 4 SPEECH ENHANCEMENT USING MULTI-BAND WIENER FILTER. In real environmental conditions the speech signal may be

CHAPTER 4 SPEECH ENHANCEMENT USING MULTI-BAND WIENER FILTER. In real environmental conditions the speech signal may be 55 CHAPTER 4 SPEECH ENHANCEMENT USING MULTI-BAND WIENER FILTER 4.1 Introducton In real envronmental condtons the speech sgnal may be supermposed by the envronmental nterference. In general, the spectrum

More information

The Wavelet Transform-Domain LMS Adaptive Filter Algorithm with Variable Step-Size

The Wavelet Transform-Domain LMS Adaptive Filter Algorithm with Variable Step-Size [ DOI:.68/IJEEE.3.3.3 Downloaded from jeee.ust.ac.r at 3:44 IRS on Wednesday ovember 4th 8 he Wavelet ransformdoman LS Adaptve Flter Algorthm wth Varable StepSze. Shams Esfand Abad*(C.A.), H. esgaran*

More information

Natural Images, Gaussian Mixtures and Dead Leaves Supplementary Material

Natural Images, Gaussian Mixtures and Dead Leaves Supplementary Material Natural Images, Gaussan Mxtures and Dead Leaves Supplementary Materal Danel Zoran Interdscplnary Center for Neural Computaton Hebrew Unversty of Jerusalem Israel http://www.cs.huj.ac.l/ danez Yar Wess

More information

ECE559VV Project Report

ECE559VV Project Report ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate

More information

A Robust Method for Calculating the Correlation Coefficient

A Robust Method for Calculating the Correlation Coefficient A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal

More information

Time-Varying Systems and Computations Lecture 6

Time-Varying Systems and Computations Lecture 6 Tme-Varyng Systems and Computatons Lecture 6 Klaus Depold 14. Januar 2014 The Kalman Flter The Kalman estmaton flter attempts to estmate the actual state of an unknown dscrete dynamcal system, gven nosy

More information

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin Proceedngs of the 007 Wnter Smulaton Conference S G Henderson, B Bller, M-H Hseh, J Shortle, J D Tew, and R R Barton, eds LOW BIAS INTEGRATED PATH ESTIMATORS James M Calvn Department of Computer Scence

More information

Average Decision Threshold of CA CFAR and excision CFAR Detectors in the Presence of Strong Pulse Jamming 1

Average Decision Threshold of CA CFAR and excision CFAR Detectors in the Presence of Strong Pulse Jamming 1 Average Decson hreshold of CA CFAR and excson CFAR Detectors n the Presence of Strong Pulse Jammng Ivan G. Garvanov and Chrsto A. Kabachev Insttute of Informaton echnologes Bulgaran Academy of Scences

More information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve

More information

Supporting Information

Supporting Information Supportng Informaton The neural network f n Eq. 1 s gven by: f x l = ReLU W atom x l + b atom, 2 where ReLU s the element-wse rectfed lnear unt, 21.e., ReLUx = max0, x, W atom R d d s the weght matrx to

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

The Concept of Beamforming

The Concept of Beamforming ELG513 Smart Antennas S.Loyka he Concept of Beamformng Generc representaton of the array output sgnal, 1 where w y N 1 * = 1 = w x = w x (4.1) complex weghts, control the array pattern; y and x - narrowband

More information

arxiv:cs.cv/ Jun 2000

arxiv:cs.cv/ Jun 2000 Correlaton over Decomposed Sgnals: A Non-Lnear Approach to Fast and Effectve Sequences Comparson Lucano da Fontoura Costa arxv:cs.cv/0006040 28 Jun 2000 Cybernetc Vson Research Group IFSC Unversty of São

More information

Parameter Estimation for Dynamic System using Unscented Kalman filter

Parameter Estimation for Dynamic System using Unscented Kalman filter Parameter Estmaton for Dynamc System usng Unscented Kalman flter Jhoon Seung 1,a, Amr Atya F. 2,b, Alexander G.Parlos 3,c, and Klto Chong 1,4,d* 1 Dvson of Electroncs Engneerng, Chonbuk Natonal Unversty,

More information

Multigradient for Neural Networks for Equalizers 1

Multigradient for Neural Networks for Equalizers 1 Multgradent for Neural Netorks for Equalzers 1 Chulhee ee, Jnook Go and Heeyoung Km Department of Electrcal and Electronc Engneerng Yonse Unversty 134 Shnchon-Dong, Seodaemun-Ku, Seoul 1-749, Korea ABSTRACT

More information

Identification of Linear Partial Difference Equations with Constant Coefficients

Identification of Linear Partial Difference Equations with Constant Coefficients J. Basc. Appl. Sc. Res., 3(1)6-66, 213 213, TextRoad Publcaton ISSN 29-434 Journal of Basc and Appled Scentfc Research www.textroad.com Identfcaton of Lnear Partal Dfference Equatons wth Constant Coeffcents

More information

Tracking with Kalman Filter

Tracking with Kalman Filter Trackng wth Kalman Flter Scott T. Acton Vrgna Image and Vdeo Analyss (VIVA), Charles L. Brown Department of Electrcal and Computer Engneerng Department of Bomedcal Engneerng Unversty of Vrgna, Charlottesvlle,

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

The Synchronous 8th-Order Differential Attack on 12 Rounds of the Block Cipher HyRAL

The Synchronous 8th-Order Differential Attack on 12 Rounds of the Block Cipher HyRAL The Synchronous 8th-Order Dfferental Attack on 12 Rounds of the Block Cpher HyRAL Yasutaka Igarash, Sej Fukushma, and Tomohro Hachno Kagoshma Unversty, Kagoshma, Japan Emal: {garash, fukushma, hachno}@eee.kagoshma-u.ac.jp

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method Appled Mathematcal Scences, Vol. 7, 0, no. 47, 07-0 HIARI Ltd, www.m-hkar.com Comparson of the Populaton Varance Estmators of -Parameter Exponental Dstrbuton Based on Multple Crtera Decson Makng Method

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010 Parametrc fractonal mputaton for mssng data analyss Jae Kwang Km Survey Workng Group Semnar March 29, 2010 1 Outlne Introducton Proposed method Fractonal mputaton Approxmaton Varance estmaton Multple mputaton

More information

IMAGE DENOISING USING NEW ADAPTIVE BASED MEDIAN FILTER

IMAGE DENOISING USING NEW ADAPTIVE BASED MEDIAN FILTER Sgnal & Image Processng : An Internatonal Journal (SIPIJ) Vol.5, No.4, August 2014 IMAGE DENOISING USING NEW ADAPTIVE BASED MEDIAN FILTER Suman Shrestha 1, 2 1 Unversty of Massachusetts Medcal School,

More information

Study on Active Micro-vibration Isolation System with Linear Motor Actuator. Gong-yu PAN, Wen-yan GU and Dong LI

Study on Active Micro-vibration Isolation System with Linear Motor Actuator. Gong-yu PAN, Wen-yan GU and Dong LI 2017 2nd Internatonal Conference on Electrcal and Electroncs: echnques and Applcatons (EEA 2017) ISBN: 978-1-60595-416-5 Study on Actve Mcro-vbraton Isolaton System wth Lnear Motor Actuator Gong-yu PAN,

More information

Adaptive Consensus Control of Multi-Agent Systems with Large Uncertainty and Time Delays *

Adaptive Consensus Control of Multi-Agent Systems with Large Uncertainty and Time Delays * Journal of Robotcs, etworkng and Artfcal Lfe, Vol., o. (September 04), 5-9 Adaptve Consensus Control of Mult-Agent Systems wth Large Uncertanty and me Delays * L Lu School of Mechancal Engneerng Unversty

More information

IMPROVED STEADY STATE ANALYSIS OF THE RECURSIVE LEAST SQUARES ALGORITHM

IMPROVED STEADY STATE ANALYSIS OF THE RECURSIVE LEAST SQUARES ALGORITHM IMPROVED STEADY STATE ANALYSIS OF THE RECURSIVE LEAST SQUARES ALGORITHM Muhammad Monuddn,2 Tareq Y. Al-Naffour 3 Khaled A. Al-Hujal 4 Electrcal and Computer Engneerng Department, Kng Abdulazz Unversty

More information

Analysis of Effort Constraint Algorithm in Active Noise Control Systems

Analysis of Effort Constraint Algorithm in Active Noise Control Systems Hndaw Publshng Corporaton EURASIP Journal on Appled Sgnal Processng Volume 26, Artcle ID 54649, Pages 1 9 DOI 1.1155/ASP/26/54649 Analyss of Effort Constrant Algorthm n Actve Nose Control Systems F. Tarngoo,

More information

Convergence Analysis of Narrowband Active Noise Equalizer System under Imperfect Secondary Path Estimation

Convergence Analysis of Narrowband Active Noise Equalizer System under Imperfect Secondary Path Estimation Convergence Analyss of Narrowband Actve Nose Equalzer System under Imperfect Secondary Path Estmaton Lang Wang, Student Member, IEEE and Woon-Seng Gan, Senor Member, IEEE Abstract---- Actve nose equalzer

More information

MLE and Bayesian Estimation. Jie Tang Department of Computer Science & Technology Tsinghua University 2012

MLE and Bayesian Estimation. Jie Tang Department of Computer Science & Technology Tsinghua University 2012 MLE and Bayesan Estmaton Je Tang Department of Computer Scence & Technology Tsnghua Unversty 01 1 Lnear Regresson? As the frst step, we need to decde how we re gong to represent the functon f. One example:

More information

1 Convex Optimization

1 Convex Optimization Convex Optmzaton We wll consder convex optmzaton problems. Namely, mnmzaton problems where the objectve s convex (we assume no constrants for now). Such problems often arse n machne learnng. For example,

More information

Comparison of Wiener Filter solution by SVD with decompositions QR and QLP

Comparison of Wiener Filter solution by SVD with decompositions QR and QLP Proceedngs of the 6th WSEAS Int Conf on Artfcal Intellgence, Knowledge Engneerng and Data Bases, Corfu Island, Greece, February 6-9, 007 7 Comparson of Wener Flter soluton by SVD wth decompostons QR and

More information

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal

More information

Topic 23 - Randomized Complete Block Designs (RCBD)

Topic 23 - Randomized Complete Block Designs (RCBD) Topc 3 ANOVA (III) 3-1 Topc 3 - Randomzed Complete Block Desgns (RCBD) Defn: A Randomzed Complete Block Desgn s a varant of the completely randomzed desgn (CRD) that we recently learned. In ths desgn,

More information

Generalized Linear Methods

Generalized Linear Methods Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

More information

An Improved multiple fractal algorithm

An Improved multiple fractal algorithm Advanced Scence and Technology Letters Vol.31 (MulGraB 213), pp.184-188 http://dx.do.org/1.1427/astl.213.31.41 An Improved multple fractal algorthm Yun Ln, Xaochu Xu, Jnfeng Pang College of Informaton

More information

De-noising Method Based on Kernel Adaptive Filtering for Telemetry Vibration Signal of the Vehicle Test Kejun ZENG

De-noising Method Based on Kernel Adaptive Filtering for Telemetry Vibration Signal of the Vehicle Test Kejun ZENG 6th Internatonal Conference on Mechatroncs, Materals, Botechnology and Envronment (ICMMBE 6) De-nosng Method Based on Kernel Adaptve Flterng for elemetry Vbraton Sgnal of the Vehcle est Kejun ZEG PLA 955

More information

COMPUTATIONALLY EFFICIENT WAVELET AFFINE INVARIANT FUNCTIONS FOR SHAPE RECOGNITION. Erdem Bala, Dept. of Electrical and Computer Engineering,

COMPUTATIONALLY EFFICIENT WAVELET AFFINE INVARIANT FUNCTIONS FOR SHAPE RECOGNITION. Erdem Bala, Dept. of Electrical and Computer Engineering, COMPUTATIONALLY EFFICIENT WAVELET AFFINE INVARIANT FUNCTIONS FOR SHAPE RECOGNITION Erdem Bala, Dept. of Electrcal and Computer Engneerng, Unversty of Delaware, 40 Evans Hall, Newar, DE, 976 A. Ens Cetn,

More information

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 ) Kangweon-Kyungk Math. Jour. 4 1996), No. 1, pp. 7 16 AN ITERATIVE ROW-ACTION METHOD FOR MULTICOMMODITY TRANSPORTATION PROBLEMS Yong Joon Ryang Abstract. The optmzaton problems wth quadratc constrants often

More information

Linear Feature Engineering 11

Linear Feature Engineering 11 Lnear Feature Engneerng 11 2 Least-Squares 2.1 Smple least-squares Consder the followng dataset. We have a bunch of nputs x and correspondng outputs y. The partcular values n ths dataset are x y 0.23 0.19

More information

x = , so that calculated

x = , so that calculated Stat 4, secton Sngle Factor ANOVA notes by Tm Plachowsk n chapter 8 we conducted hypothess tests n whch we compared a sngle sample s mean or proporton to some hypotheszed value Chapter 9 expanded ths to

More information

Tornado and Luby Transform Codes. Ashish Khisti Presentation October 22, 2003

Tornado and Luby Transform Codes. Ashish Khisti Presentation October 22, 2003 Tornado and Luby Transform Codes Ashsh Khst 6.454 Presentaton October 22, 2003 Background: Erasure Channel Elas[956] studed the Erasure Channel β x x β β x 2 m x 2 k? Capacty of Noseless Erasure Channel

More information

Chapter 12 Analysis of Covariance

Chapter 12 Analysis of Covariance Chapter Analyss of Covarance Any scentfc experment s performed to know somethng that s unknown about a group of treatments and to test certan hypothess about the correspondng treatment effect When varablty

More information

Boostrapaggregating (Bagging)

Boostrapaggregating (Bagging) Boostrapaggregatng (Baggng) An ensemble meta-algorthm desgned to mprove the stablty and accuracy of machne learnng algorthms Can be used n both regresson and classfcaton Reduces varance and helps to avod

More information

A Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach

A Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach A Bayes Algorthm for the Multtask Pattern Recognton Problem Drect Approach Edward Puchala Wroclaw Unversty of Technology, Char of Systems and Computer etworks, Wybrzeze Wyspanskego 7, 50-370 Wroclaw, Poland

More information

Department of Statistics University of Toronto STA305H1S / 1004 HS Design and Analysis of Experiments Term Test - Winter Solution

Department of Statistics University of Toronto STA305H1S / 1004 HS Design and Analysis of Experiments Term Test - Winter Solution Department of Statstcs Unversty of Toronto STA35HS / HS Desgn and Analyss of Experments Term Test - Wnter - Soluton February, Last Name: Frst Name: Student Number: Instructons: Tme: hours. Ads: a non-programmable

More information

Errors for Linear Systems

Errors for Linear Systems Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch

More information

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system Transfer Functons Convenent representaton of a lnear, dynamc model. A transfer functon (TF) relates one nput and one output: x t X s y t system Y s The followng termnology s used: x y nput output forcng

More information

Radar Trackers. Study Guide. All chapters, problems, examples and page numbers refer to Applied Optimal Estimation, A. Gelb, Ed.

Radar Trackers. Study Guide. All chapters, problems, examples and page numbers refer to Applied Optimal Estimation, A. Gelb, Ed. Radar rackers Study Gude All chapters, problems, examples and page numbers refer to Appled Optmal Estmaton, A. Gelb, Ed. Chapter Example.0- Problem Statement wo sensors Each has a sngle nose measurement

More information

An identification algorithm of model kinetic parameters of the interfacial layer growth in fiber composites

An identification algorithm of model kinetic parameters of the interfacial layer growth in fiber composites IOP Conference Seres: Materals Scence and Engneerng PAPER OPE ACCESS An dentfcaton algorthm of model knetc parameters of the nterfacal layer growth n fber compostes o cte ths artcle: V Zubov et al 216

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Experment-I MODULE VII LECTURE - 3 ANALYSIS OF COVARIANCE Dr Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur Any scentfc experment s performed

More information

Logistic Regression. CAP 5610: Machine Learning Instructor: Guo-Jun QI

Logistic Regression. CAP 5610: Machine Learning Instructor: Guo-Jun QI Logstc Regresson CAP 561: achne Learnng Instructor: Guo-Jun QI Bayes Classfer: A Generatve model odel the posteror dstrbuton P(Y X) Estmate class-condtonal dstrbuton P(X Y) for each Y Estmate pror dstrbuton

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

CHAPTER 5 MINIMAX MEAN SQUARE ESTIMATION

CHAPTER 5 MINIMAX MEAN SQUARE ESTIMATION 110 CHAPTER 5 MINIMAX MEAN SQUARE ESTIMATION 5.1 INTRODUCTION The problem of parameter estmaton n lnear model s pervasve n sgnal processng and communcaton applcatons. It s often common to restrct attenton

More information

Chapter 13: Multiple Regression

Chapter 13: Multiple Regression Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to

More information

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models Computaton of Hgher Order Moments from Two Multnomal Overdsperson Lkelhood Models BY J. T. NEWCOMER, N. K. NEERCHAL Department of Mathematcs and Statstcs, Unversty of Maryland, Baltmore County, Baltmore,

More information

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

More information

Regularized Discriminant Analysis for Face Recognition

Regularized Discriminant Analysis for Face Recognition 1 Regularzed Dscrmnant Analyss for Face Recognton Itz Pma, Mayer Aladem Department of Electrcal and Computer Engneerng, Ben-Guron Unversty of the Negev P.O.Box 653, Beer-Sheva, 845, Israel. Abstract Ths

More information

VQ widely used in coding speech, image, and video

VQ widely used in coding speech, image, and video at Scalar quantzers are specal cases of vector quantzers (VQ): they are constraned to look at one sample at a tme (memoryless) VQ does not have such constrant better RD perfomance expected Source codng

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 31 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 6. Rdge regresson The OLSE s the best lnear unbased

More information

Basic Statistical Analysis and Yield Calculations

Basic Statistical Analysis and Yield Calculations October 17, 007 Basc Statstcal Analyss and Yeld Calculatons Dr. José Ernesto Rayas Sánchez 1 Outlne Sources of desgn-performance uncertanty Desgn and development processes Desgn for manufacturablty A general

More information

COMPARING NOISE REMOVAL IN THE WAVELET AND FOURIER DOMAINS

COMPARING NOISE REMOVAL IN THE WAVELET AND FOURIER DOMAINS COMPARING NOISE REMOVAL IN THE WAVELET AND FOURIER DOMAINS Robert J. Barsant, and Jordon Glmore Department of Electrcal and Computer Engneerng The Ctadel Charleston, SC, 29407 e-mal: robert.barsant@ctadel.edu

More information

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations Physcs 171/271 - Chapter 9R -Davd Klenfeld - Fall 2005 9 Dervaton of Rate Equatons from Sngle-Cell Conductance (Hodgkn-Huxley-lke) Equatons We consder a network of many neurons, each of whch obeys a set

More information

829. An adaptive method for inertia force identification in cantilever under moving mass

829. An adaptive method for inertia force identification in cantilever under moving mass 89. An adaptve method for nerta force dentfcaton n cantlever under movng mass Qang Chen 1, Mnzhuo Wang, Hao Yan 3, Haonan Ye 4, Guola Yang 5 1,, 3, 4 Department of Control and System Engneerng, Nanng Unversty,

More information

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control An Upper Bound on SINR Threshold for Call Admsson Control n Multple-Class CDMA Systems wth Imperfect ower-control Mahmoud El-Sayes MacDonald, Dettwler and Assocates td. (MDA) Toronto, Canada melsayes@hotmal.com

More information

Pulse Coded Modulation

Pulse Coded Modulation Pulse Coded Modulaton PCM (Pulse Coded Modulaton) s a voce codng technque defned by the ITU-T G.711 standard and t s used n dgtal telephony to encode the voce sgnal. The frst step n the analog to dgtal

More information

Design and Optimization of Fuzzy Controller for Inverse Pendulum System Using Genetic Algorithm

Design and Optimization of Fuzzy Controller for Inverse Pendulum System Using Genetic Algorithm Desgn and Optmzaton of Fuzzy Controller for Inverse Pendulum System Usng Genetc Algorthm H. Mehraban A. Ashoor Unversty of Tehran Unversty of Tehran h.mehraban@ece.ut.ac.r a.ashoor@ece.ut.ac.r Abstract:

More information

Neuro-Adaptive Design - I:

Neuro-Adaptive Design - I: Lecture 36 Neuro-Adaptve Desgn - I: A Robustfyng ool for Dynamc Inverson Desgn Dr. Radhakant Padh Asst. Professor Dept. of Aerospace Engneerng Indan Insttute of Scence - Bangalore Motvaton Perfect system

More information

Negative Binomial Regression

Negative Binomial Regression STATGRAPHICS Rev. 9/16/2013 Negatve Bnomal Regresson Summary... 1 Data Input... 3 Statstcal Model... 3 Analyss Summary... 4 Analyss Optons... 7 Plot of Ftted Model... 8 Observed Versus Predcted... 10 Predctons...

More information

JAB Chain. Long-tail claims development. ASTIN - September 2005 B.Verdier A. Klinger

JAB Chain. Long-tail claims development. ASTIN - September 2005 B.Verdier A. Klinger JAB Chan Long-tal clams development ASTIN - September 2005 B.Verder A. Klnger Outlne Chan Ladder : comments A frst soluton: Munch Chan Ladder JAB Chan Chan Ladder: Comments Black lne: average pad to ncurred

More information

One-sided finite-difference approximations suitable for use with Richardson extrapolation

One-sided finite-difference approximations suitable for use with Richardson extrapolation Journal of Computatonal Physcs 219 (2006) 13 20 Short note One-sded fnte-dfference approxmatons sutable for use wth Rchardson extrapolaton Kumar Rahul, S.N. Bhattacharyya * Department of Mechancal Engneerng,

More information

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng

More information

Hopfield Training Rules 1 N

Hopfield Training Rules 1 N Hopfeld Tranng Rules To memorse a sngle pattern Suppose e set the eghts thus - = p p here, s the eght beteen nodes & s the number of nodes n the netor p s the value requred for the -th node What ll the

More information

Novel Pre-Compression Rate-Distortion Optimization Algorithm for JPEG 2000

Novel Pre-Compression Rate-Distortion Optimization Algorithm for JPEG 2000 Novel Pre-Compresson Rate-Dstorton Optmzaton Algorthm for JPEG 2000 Yu-We Chang, Hung-Ch Fang, Chung-Jr Lan, and Lang-Gee Chen DSP/IC Desgn Laboratory, Graduate Insttute of Electroncs Engneerng Natonal

More information

1 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations

1 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations Physcs 171/271 -Davd Klenfeld - Fall 2005 (revsed Wnter 2011) 1 Dervaton of Rate Equatons from Sngle-Cell Conductance (Hodgkn-Huxley-lke) Equatons We consder a network of many neurons, each of whch obeys

More information

FAST CONVERGENCE ADAPTIVE MMSE RECEIVER FOR ASYNCHRONOUS DS-CDMA SYSTEMS

FAST CONVERGENCE ADAPTIVE MMSE RECEIVER FOR ASYNCHRONOUS DS-CDMA SYSTEMS Électronque et transmsson de l nformaton FAST CONVERGENCE ADAPTIVE MMSE RECEIVER FOR ASYNCHRONOUS DS-CDMA SYSTEMS CĂLIN VLĂDEANU, CONSTANTIN PALEOLOGU 1 Key words: DS-CDMA, MMSE adaptve recever, Least

More information

Homework Assignment 3 Due in class, Thursday October 15

Homework Assignment 3 Due in class, Thursday October 15 Homework Assgnment 3 Due n class, Thursday October 15 SDS 383C Statstcal Modelng I 1 Rdge regresson and Lasso 1. Get the Prostrate cancer data from http://statweb.stanford.edu/~tbs/elemstatlearn/ datasets/prostate.data.

More information

This column is a continuation of our previous column

This column is a continuation of our previous column Comparson of Goodness of Ft Statstcs for Lnear Regresson, Part II The authors contnue ther dscusson of the correlaton coeffcent n developng a calbraton for quanttatve analyss. Jerome Workman Jr. and Howard

More information

8 Derivation of Network Rate Equations from Single- Cell Conductance Equations

8 Derivation of Network Rate Equations from Single- Cell Conductance Equations Physcs 178/278 - Davd Klenfeld - Wnter 2015 8 Dervaton of Network Rate Equatons from Sngle- Cell Conductance Equatons We consder a network of many neurons, each of whch obeys a set of conductancebased,

More information

Lecture 10 Support Vector Machines II

Lecture 10 Support Vector Machines II Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed

More information

Multi-user Detection Based on Weight approaching particle filter in Impulsive Noise

Multi-user Detection Based on Weight approaching particle filter in Impulsive Noise Internatonal Symposum on Computers & Informatcs (ISCI 2015) Mult-user Detecton Based on Weght approachng partcle flter n Impulsve Nose XIAN Jn long 1, a, LI Sheng Je 2,b 1 College of Informaton Scence

More information

Chapter - 2. Distribution System Power Flow Analysis

Chapter - 2. Distribution System Power Flow Analysis Chapter - 2 Dstrbuton System Power Flow Analyss CHAPTER - 2 Radal Dstrbuton System Load Flow 2.1 Introducton Load flow s an mportant tool [66] for analyzng electrcal power system network performance. Load

More information

A Particle Filter Algorithm based on Mixing of Prior probability density and UKF as Generate Importance Function

A Particle Filter Algorithm based on Mixing of Prior probability density and UKF as Generate Importance Function Advanced Scence and Technology Letters, pp.83-87 http://dx.do.org/10.14257/astl.2014.53.20 A Partcle Flter Algorthm based on Mxng of Pror probablty densty and UKF as Generate Importance Functon Lu Lu 1,1,

More information

CIS526: Machine Learning Lecture 3 (Sept 16, 2003) Linear Regression. Preparation help: Xiaoying Huang. x 1 θ 1 output... θ M x M

CIS526: Machine Learning Lecture 3 (Sept 16, 2003) Linear Regression. Preparation help: Xiaoying Huang. x 1 θ 1 output... θ M x M CIS56: achne Learnng Lecture 3 (Sept 6, 003) Preparaton help: Xaoyng Huang Lnear Regresson Lnear regresson can be represented by a functonal form: f(; θ) = θ 0 0 +θ + + θ = θ = 0 ote: 0 s a dummy attrbute

More information

Suppose that there s a measured wndow of data fff k () ; :::; ff k g of a sze w, measured dscretely wth varable dscretzaton step. It s convenent to pl

Suppose that there s a measured wndow of data fff k () ; :::; ff k g of a sze w, measured dscretely wth varable dscretzaton step. It s convenent to pl RECURSIVE SPLINE INTERPOLATION METHOD FOR REAL TIME ENGINE CONTROL APPLICATIONS A. Stotsky Volvo Car Corporaton Engne Desgn and Development Dept. 97542, HA1N, SE- 405 31 Gothenburg Sweden. Emal: astotsky@volvocars.com

More information

Low Complexity Soft-Input Soft-Output Hamming Decoder

Low Complexity Soft-Input Soft-Output Hamming Decoder Low Complexty Soft-Input Soft-Output Hammng Der Benjamn Müller, Martn Holters, Udo Zölzer Helmut Schmdt Unversty Unversty of the Federal Armed Forces Department of Sgnal Processng and Communcatons Holstenhofweg

More information

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests Smulated of the Cramér-von Mses Goodness-of-Ft Tests Steele, M., Chaselng, J. and 3 Hurst, C. School of Mathematcal and Physcal Scences, James Cook Unversty, Australan School of Envronmental Studes, Grffth

More information

T E C O L O T E R E S E A R C H, I N C.

T E C O L O T E R E S E A R C H, I N C. T E C O L O T E R E S E A R C H, I N C. B rdg n g En g neern g a nd Econo mcs S nce 1973 THE MINIMUM-UNBIASED-PERCENTAGE ERROR (MUPE) METHOD IN CER DEVELOPMENT Thrd Jont Annual ISPA/SCEA Internatonal Conference

More information

SYNTHETIC aperture radar (SAR) uses the relative motion

SYNTHETIC aperture radar (SAR) uses the relative motion 512 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 3, NO. 4, OCTOBER 2006 SAR Mnmum-Entropy Autofocus Usng an Adaptve-Order Polynomal Model Junfeng Wang and Xngzhao Lu Abstract A new algorthm s presented

More information

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM Internatonal Conference on Ceramcs, Bkaner, Inda Internatonal Journal of Modern Physcs: Conference Seres Vol. 22 (2013) 757 761 World Scentfc Publshng Company DOI: 10.1142/S2010194513010982 FUZZY GOAL

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers Psychology 282 Lecture #24 Outlne Regresson Dagnostcs: Outlers In an earler lecture we studed the statstcal assumptons underlyng the regresson model, ncludng the followng ponts: Formal statement of assumptons.

More information

Research on Modified Root-MUSIC Algorithm of DOA Estimation Based on Covariance Matrix Reconstruction

Research on Modified Root-MUSIC Algorithm of DOA Estimation Based on Covariance Matrix Reconstruction Sensors & ransducers, Vol. 78, Issue 9, September 04, pp. 4-8 Sensors & ransducers 04 by IFSA Publshng, S. L. http://www.sensorsportal.com Research on Modfed Root-MUSIC Algorthm of DOA Estmaton Based on

More information

COEFFICIENT DIAGRAM: A NOVEL TOOL IN POLYNOMIAL CONTROLLER DESIGN

COEFFICIENT DIAGRAM: A NOVEL TOOL IN POLYNOMIAL CONTROLLER DESIGN Int. J. Chem. Sc.: (4), 04, 645654 ISSN 097768X www.sadgurupublcatons.com COEFFICIENT DIAGRAM: A NOVEL TOOL IN POLYNOMIAL CONTROLLER DESIGN R. GOVINDARASU a, R. PARTHIBAN a and P. K. BHABA b* a Department

More information

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law:

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law: CE304, Sprng 2004 Lecture 4 Introducton to Vapor/Lqud Equlbrum, part 2 Raoult s Law: The smplest model that allows us do VLE calculatons s obtaned when we assume that the vapor phase s an deal gas, and

More information

Uncertainty and auto-correlation in. Measurement

Uncertainty and auto-correlation in. Measurement Uncertanty and auto-correlaton n arxv:1707.03276v2 [physcs.data-an] 30 Dec 2017 Measurement Markus Schebl Federal Offce of Metrology and Surveyng (BEV), 1160 Venna, Austra E-mal: markus.schebl@bev.gv.at

More information