OVERCOMPLETE BLIND SOURCE SEPARATION BY COMBINING ICA AND BINARY TIME-FREQUENCY MASKING
|
|
- Rosamund Potter
- 6 years ago
- Views:
Transcription
1 OVERCOMPLETE BLIND SOURCE SEPARATION BY COMBINING ICA AND BINARY TIME-FREQUENCY MASKING Michael Syskid Pederse 1,2, DeLiag Wag 3, Ja Larse 1 ad Ulrik Kjems 2 1 Iformatics ad Mathematical Modellig, Techical Uiversity of Demark Richard Peterses Plads, Buildig 321, DK-28 Kgs. Lygby, Demark 2 Otico A/S, Stradveje 58, DK-29 Hellerup, Demark 3 Departmet of Computer Sciece ad Egieerig & Ceter for Cogitive Sciece, The Ohio State Uiversity, Columbus, OH , USA ABSTRACT A limitatio i may source separatio tasks is that the umber of source sigals has to be kow i advace. Further, i order to achieve good performace, the umber of sources caot exceed the umber of sesors. I may real-world applicatios these limitatios are too strict. We propose a ovel method for overcomplete blid source separatio. Two powerful source separatio techiques have bee combied, idepedet compoet aalysis ad biary time-frequecy maskig. Hereby, it is possible to iteratively extract each speech sigal from the mixture. By usig merely two microphoes we ca separate up to six mixed speech sigals uder aechoic coditios. The umber of source sigals is ot assumed to be kow i advace. It is also possible to maitai the extracted sigals as stereo sigals. 1. INTRODUCTION Blid source separatio (BSS) addresses the problem of recoverig N ukow source sigals s() = [s 1(),..., s N()] T from M recorded mixtures x() = [x 1(),..., x M()] T of the source sigals. The term blid refers to that oly the recorded mixtures are kow. A importat applicatio for BSS is separatio of speech sigals. The recorded mixtures are assumed to be liear superpositios of the source sigals, i.e. x() = As() + ν(), (1) where A is a M N mixig matrix ad deotes the discrete time idex. ν() is additioal oise. A method to retrieve the origial sigals up to a arbitrary permutatio ad scalig is idepedet compoet aalysis (ICA) [1]. I ICA, the mai assumptio is that the source sigals are idepedet. By applyig ICA, a estimate y() of the source sigals ca be obtaied by fidig a (pseudo)iverse W of the mixig matrix so that y() = Wx(). (2) May methods require that the umber of source sigals is kow i advace. Aother drawback of most of these methods is that the umber of source sigals is assumed ot to exceed the umber of microphoes, i.e. M N. Eve if the mixig process A is kow, it is ot ivertible, ad i geeral, the idepedet compoets caot be recovered exactly [1]. I the case of more sources tha sesors, the overcomplete/uderdetermied case, successful separatio ofte relies o the assumptio that the source sigals are sparsely distributed - either i the time domai, i the frequecy domai or i the time-frequecy (T-F) domai [2], [3], [4], [5]. If the source sigals do ot overlap i the time-frequecy domai, high-quality recostructio could be obtaied [4]. However, there is overlap betwee the source sigals. I this case, good separatio ca still be obtaied by applyig a biary time-frequecy mask to the mixture [3], [4]. I computatioal auditory scee aalysis, the techique of T-F maskig has bee commoly used for years (see e.g. [6]). Here, source separatio is based o orgaizatioal cues from auditory scee aalysis [7]. More recetly the techique has also become popular i blid source separatio, where separatio is based o o-overlappig sources i the T-F domai [8]. T-F maskig is applicable to source separatio/ segregatio usig oe microphoe [6], [9] or more tha oe microphoe [3], [4]. T-F maskig ca be applied as a biary mask. For a biary mask, each T-F uit is either weighted by oe or by zero. I order to reduce musical oise, more smooth masks may also be applied [1]. A advatage of usig a biary mask is that oly a biary decisio has to be made [11]. Such a decisio ca be based o, e.g., clusterig [3], [4], [8], or directio-of-arrival [12]. ICA has bee used i differet combiatios with the biary mask. I [12], separatio is performed by removig sigals by maskig N M sigals ad afterwards applyig ICA i order to separate the remaiig M sigals. ICA has also bee used the other way aroud. I [13], it has bee applied to separate two sigals by usig two microphoes. Based o the ICA outputs, T- F masks are estimated ad a mask is applied to each of the ICA outputs i order to improve the sigal to oise ratio. I this paper, a ovel method for separatig a arbitrary umber of speech sigals is proposed. Based o the output of a square (2 2) ICA algorithm ad biary T-F masks, this method iteratively segregates sigals from a mixture util a estimate of each sigal is obtaied. 2. GEOMETRICAL INTERPRETATION OF INSTANTANEOUS ICA We assume that there is a ukow umber of acoustical source sigals but oly two microphoes. It is assumed that each source sigal arrives from a certai directio ad o reflectios occur, i.e. a aechoic eviromet. I order to keep the problem simple, the source sigals are mixed by a istataeous mixig matrix as i eq. (1). Due to delays betwee the microphoes, istataeous ICA with a real-valued mixig matrix usually is ot applica-
2 r 1 (θ) r 2 (θ) FRm RUf 9 CNf 6 ITf dB 1dB db 1dB UKm NLm 18 2dB 1dB db 1dB Fig. 1. The two directioal microphoe resposes are show as fuctio of the directio θ. Table 1. The six speech sigals. All speakers use raised voice as if they were speakig i a oisy eviromet. Abbreviatio CNf NLm FRm ITf UKm RUf Descriptio Female speech i Chiese Male speech i Dutch Male speech i Frech Female speech i Italia Male speech i Eglish Female speech i Russia ble to sigals recorded at a array of microphoes, but if the microphoes are placed at exact same locatio ad the microphoes have differet resposes for differet directios, the separatio of delayed sources ca be approximated by the istataeous model [14]. Hereby, a combiatio of microphoe gais correspod to a certai directioal patter. Therefore, two directioal microphoe resposes are used. The two microphoe resposes are chose as fuctios of the directio θ as r 1(θ) = cos(θ) ad r 2(θ) = 1.5 cos(θ), respectively. The two microphoe resposes are show i figure 1. It is possible to make two such directioal patters by addig ad subtractig omidirectioal sigals from two microphoes placed closely together. Hece, the mixig system is give by A(θ) = r1(θ 1) r 1(θ N) r 2(θ 1) r 2(θ N). (3) Differet speech sigals are used as source sigals. The used sigals are sampled with a samplig frequecy of 1 khz ad the duratio of each sigal is 5 s. The speech sigals are show i table More sources tha sesors Now cosider the case where N (M = 2). Whe there are oly two mixed sigals, a stadard ICA algorithm oly has two Fig. 2. The polar plots show the gai for differet directios. ICA is applied with two sesors ad six sources. The two dots at the periphery show the ull directios. The lies poitig out from the origi deote the true directio of the speech sources. The threeletter abbreviatios (see table 1) idetifies the differet speech sigals which have bee used. As it ca be see from the figure, the ICA solutio teds to place the ull towards sources spatially close to each other. Therefore, each of the two outputs is a group of sigals spatially close to each other. output sigals y() = [y 1(), y 2()] T. Sice the umber of separated sigals obtaied by (2) is smaller tha the umber of source sigals, y does ot cotai the separated sigals. Istead y is aother liear superpositio of each of the source sigals, where the weights are give by G = WA istead of just A as i (1). Hereby, G just correspods to aother weightig depedig o θ. These weights make y 1() ad y 2() as idepedet as possible. This is illustrated i figure 2. A implemetatio of the ifomax ICA algorithm [15] has bee used. The BGFS method has bee used for optimizatio [16] 1. The figure shows the two estimated spatial resposes from G(θ) i the overdetermied case. The respose of the m th output is give by w T ma(θ), where w m is the separatio vector from the m th output ad a(θ) is the mixig vector for the arrival directio θ [17]. By varyig θ over all possible directios, directivity patters ca be created as show i figure 2. The estimated ull placemet is illustrated by the two roud dots placed at the periphery of the polar plot. The lies poitig out from the origi illustrate the correct directio of the source sigals. Here, the sources are uiformly distributed i the iterval [ θ 18 ]. As it ca be see, the ulls do ot cacel sigle sources out. Rather, a ull is placed at a directio poitig towards a group of sources which are spatially close to each other. Here, it ca be see that the first output, y 1(), the sigals NLm ad FRm are domiatig ad i the secod output, y 2(), the sigals UKm, ITf ad CNf are domiatig. The sixth sigal, RUf exists i both outputs. This ew weightig of the sigals ca be used to estimate biary masks. 1 Matlab toolbox available from toolbox/ica/
3 3. BLIND SOURCE EXTRACTION WITH ICA AND BINARY MASKING A flowchart for the algorithm is give i figure 3. As described i the previous sectio, a two-iput-two-output ICA algorithm is applied to the iput mixtures, disregardig the umber of source sigals that actually exist i the mixture. The two output sigals are arbitrarily scaled. The scalig is fixed by usig kowledge about the microphoe resposes. Hereby, the two ull directios ca be foud. The two output sigals are scaled such that where oe directioal respose has a ull, the other respose has a uit gai. The two re-scaled output sigals, ŷ 1() ad ŷ 2() are trasformed ito the frequecy domai e.g. by use of the Short-Time Fourier Trasform STFT so that two spectrograms are obtaied: ŷ 1 Y 1(ω,t) (4) ŷ 2 Y 2(ω,t), (5) where ω deotes the frequecy ad t is the time idex. The biary masks are the determied by for each T-F uit comparig the amplitudes of the two spectrograms: BM1(ω, t) = τ Y 1(ω, t) > Y 2(ω,t) (6) BM2(ω, t) = τ Y 2(ω, t) > Y 1(ω,t), (7) where τ is a threshold. Next, each of the the two biary masks is applied to the origial mixtures i the T-F domai, ad by this oliear processig, some of the speech sigals are removed by oe of the masks while other speakers are removed by the other mask. After the masks have bee applied to the sigals, they are recostructed i the time domai by the iverse STFT. If there is oly a sigle sigal left i the masked output, defied by the selectio criteria i sectio 3.1, i.e. all but oe speech sigal have bee masked, this sigal has bee extracted from the mixture ad it is saved. If there are more tha oe sigal left i the masked outputs, ICA is applied to the two masked sigals agai ad a ew set of masks are created based o (6), (7) ad the previous masks. The use of the previous mask esures that T-F uits that have bee removed from the mixture are ot reitroduced by the ext mask. This is doe by a elemet-wise multiplicatio betwee the previous mask ad the ew mask. This iterative procedure is followed util all masked outputs cosist of oly a sigle speech sigal. Notice, the output sigals are maitaied as two sigals. Stereo sigals created with directioal microphoes placed at the same locatio with a agle betwee the directioal patters of 9 (here 18 ) are termed XY-stereo Selectio criterio Further processig o a pair of masked sigals should be avoided i two cases. If all but oe sigal have bee removed or if too much has bee removed so that there is o sigal left after applyig the mask. The decisios are based o the eigevalues of the covariace matrix betwee the masked sesor sigals. The covariace matrix is calculated as R = T, (8) where deotes the expectatio with respect to the whole sigal, ad is the two time domai sigals of which the biary mask has bee applied. If oly cotais oe sigal, the covariace matrix is sigular, ad the smallest eigevalue λ mi is approximately equal to zero [18]. Sice parts of the other sigals may remai after maskig, the smallest eigevalue is equal to the oise variace Cotiue x Iitializatio 1 x2 xˆ 1x ˆy 1 ICA + scalig Estimatio of the two biary masks BM 1 BM 2 Apply to origial microphoe sigals Selectio criteria Fial stereo sigal Iput sigal buffer xˆ 2 x x1 x2 x1 x2 Apply to origial microphoe sigals Stop Stop ˆy 2 Selectio criteria Fial stereo sigal Cotiue Fig. 3. Flowchart showig the mai steps of the proposed algorithm. From the output of the ICA algorithm, biary masks are estimated. The biary masks are applied to the origial sigals which agai are processed through the ICA step. Every time the output from oe of the biary masks is detected as a sigle sigal, the sigal is stored. The iterative procedure stops whe all outputs oly cosist of a sigle sigal. of these remaiig sigals. Therefore, if λ mi is smaller tha a certai oise threshold τ λmi, it is assumed that there is less tha two sigals ad o further processig is ecessary. I order to discrimiate betwee zero or oe sigal, the largest eigevalue λ max is cosidered. If λ max is smaller tha a certai threshold τ λmax, the output is cosidered of such a bad quality that the sigal should be throw away Fidig the remaiig sigals Sice some sigals may have bee removed by both masks, all T-F uits that have ot bee assiged the value 1 are used to create a remaiig mask, ad the procedure is applied to the mixture sigal of which the remaiig mask is applied, to esure that all sigals are estimated. Notice, this step has bee omitted from figure EVALUATION The algorithm described above has bee implemeted ad evaluated with mixtures of the six sigals from table 1. For the STFT,
4 a FFT legth of 248 has bee used. This gives a frequecy resolutio of 125 frequecy uits. A Haig widow with a legth of 512 samples has bee applied to the FFT sigal ad the frame shift is 256 samples. A high frequecy resolutio is foud to be ecessary i order to obtai good performace. The samplig frequecy of the speech sigals is 1 khz. The three thresholds τ, τ λmi ad τ λmax have bee foud from iitial experimets. I the ICA step, the separatio matrix is iitialized by the idetity matrix, i.e. W = I. I order to test robustess, W was also iitialized with a radom matrix with values uiformly distributed over the iterval [,1]. The differet iitializatio did ot affect the result. Whe usig a biary mask, it is ot possible to recostruct the speech sigal as if it was recorded i the absece of the iterferig sigals, because the sigals partly overlap. Therefore, as a computatioal goal for source separatio, the ideal biary mask has bee suggested [11]. The ideal biary mask for a sigal is foud for each T-F uit by comparig the eergy of the desired sigal to the eergy of all the iterferig sigals. Wheever the sigal eergy is highest, the T-F uit is assiged the value 1 ad wheever the iterferig sigals have more eergy, the T-F uit is assiged the value. As i [9], for each of the separated sigals, the percetage of eergy loss P EL ad the percetage of oise residue P NR are calculated: P EL = P NR = e 2 1() I 2 () e 2 2() (9) O 2 (), (1) where O() is the estimated sigal, ad I() is the recorded mixture resythesized after applyig the ideal biary mask. e 1() deotes the sigal preset i I() but abset i O() ad e 2() deotes the sigal preset i O() but abset i I(). Also the sigal to oise ratio (SNR) is foud. Here the SNR is defied usig the resythesized speech from the ideal biary mask as the groud truth I 2 () SNR = 1log 1. (11) (I() O()) 2 The algorithm has bee applied to mixtures cosistig of up to six sigals. I all mixig situatios, the sigals have bee uiformly distributed i the iterval [ θ 18 ]. The separatio results are show i figure 4 ad i table 2. Two ideal biary masks have bee foud oe for each microphoe sigal. I all cases, all the sigals have bee segregated from the mixture. I most cases also the correct umber of sigals is estimated. Oly i the case of three mixtures, oe of the source sigals is estimated twice. The double extractio is caused by the selectio criteria. Based o the chose thresholds, the selectio criteria i some cases allows a sigal to be extracted more tha oce. I the case of the six mixtures from figure 2, the six estimated biary masks are show i figure 5 alog with the estimated ideal biary masks from each of the two microphoe sigals. The iput SNR (SNR i) is show i figure 4 too. The SNR i is the ratio betwee the desired sigal ad the oise i the recorded mixtures. The separatio quality decreases whe the umber of sigals is Average SNR [db] Number of sources SNR i SNR Fig. 4. The sigal to oise ratio as fuctio of the umber of source sigals. The average SNR for the mixtures before separatio (SNR i) is show as well as the average SNR after separatio calculated by eq. (11). I the case of three sigals, the icorrectly estimated sigal is igored (see table 2). icreased. This is expected because whe the umber of mixed sigals is icreased, the mixtures become less sparse. Radom distributios of the source directios as well as more tha six sigals have also bee examied. Here, i geeral, ot all the sources are separated from each other. If the arrival agles betwee sigals are too arrow, these sigals may be detected as a sigle sigal, ad they are ot separated. Listeig tests validate the separatio results. This method differs from previous methods which use a biary mask ad two microphoes [3], [4]. I [3], biaural cues have bee applied for separatio, i.e iteraural time ad itesity differeces. I [4], the separatio is likewise based o amplitude ad time differece of each source. Here separatio is based o clusterig of T-F uits that have similar amplitude ad phase properties. I our approach too, separatio ca oly be achieved if the source sigals have differet spatial positios, but the separatio criterio is based o idepedece betwee the source sigals. 5. CONCLUDING REMARKS A ovel method of blid source separatio of has bee described. Based o sparseess ad idepedece, the method iteratively extracts all the speech sigals without kowig the sigals i advace. A advatage of this method is that stereo sigals are maitaied through the processig. So far, the method has bee applied to successful separatio of up to six speech sigals uder aechoic coditios by use of two microphoes. Future work will iclude separatio of mixtures i reverberat eviromet, a more blid solutio of the scalig problem, ad improved techiques for the stoppig criteria based o detectio of a sigle sigal. Alterative to usig a liear frequecy scale, a frequecy scale that models the auditory system more accurately could be used, because a auditory-based frot-ed is reported to be more robust tha a Fourier-based aalysis i the presece of backgroud iterferece [9]. The use of more tha two sesors could also be ivestigated. By usig more tha two sesors, a better resolutio ca be obtaied
5 (a) (c) (e) Microphoe 1 ideal biary mask for CNf Microphoe 2 ideal biary mask for CNf Estimated biary mask for CNf Microphoe 1 ideal biary mask for FRm Microphoe 2 ideal biary mask for FRm Estimated biary mask for FRm Microphoe 1 ideal biary mask for UKm Microphoe 2 ideal biary mask for UKm Estimated biary mask for UKm (b) Microphoe 1 ideal biary mask for NLm Microphoe 2 ideal biary mask for NLm Estimated biary mask for NLm (d) Microphoe 1 ideal biary mask for ITf (f) Microphoe 2 ideal biary mask for ITf Estimated biary mask for ITf Microphoe 1 ideal biary mask for RUf Microphoe 2 ideal biary mask for RUf Estimated biary mask for RUf Fig. 5. For a mixture of 6 mixed speech sigals, biary masks have bee estimated for each of the 6 speech sigals. The black areas correspod to the mask value 1 ad the white areas correspod to the mask value. The results are show together with the calculated ideal biary masks of each of the two microphoe sigals. The sigals (a) (f) appear i the order which they were extracted from the mixture. The first three sigals (a) (c) were extracted after two iteratios, the ext two sigals (d), (e) were extracted after three iteratios. The last sigal (f) was extracted from the remaiig mask as described i sectio 3.2.
6 Table 2. Separatio results. Mixtures cosistig from two up to six sigals have bee separated from each other successfully. I most cases, the correct umber of sources has bee extracted. Oly i the case of three source sigals, oe of the sigals has bee estimated twice. Here the average performace has bee calculated with( ) ad without the extra sigal. The sigals appear i the order which they were extracted from the mixture. Separated Microphoe 1 Microphoe 2 Sigal P EL(%) P NR(%) P EL(%) P NR(%) UKm FRm Average NLm CNf CNf RUf Average Average CNf RUf FRm UKm Average RUf NLm FRm ITf CNf Average CNf NLm FRm ITf UKm RUf Average ad ambiguous arrival agles may be avoided. Also applicatios for other types of sparse sigals could be examied. 6. ACKNOWLEDGEMENTS The work was performed while M.S.P. was a visitig scholar at The Ohio State Uiversity Departmet of Computer Sciece ad Egieerig. M.S.P was supported by the Otico Foudatio. M.S.P ad J.L are partly also supported by the Europea Commissio through the sixth framework IST Network of Excellece: Patter Aalysis, Statistical Modellig ad Computatioal Learig (PASCAL), cotract o D.L.W was supported i part by a AFOSR grat (FA ) ad a AFRL grat (FA ). 7. REFERENCES [1] A. Hyvärie, J. Karhue, ad E. Oja, Idepedet Compoet Aalysis, Wiley, 21. [2] P. Bofill ad M. Zibulevsky, Blid separatio of more sources tha mixtures usig sparsity of their short-time fourier trasform, i Proc. ICA 2, 2, pp [3] N. Roma, D. L. Wag, ad G. J. Brow, Speech segregatio based o soud localizatio, J. Acoust. Soc. Amer., vol. 114, o. 4, pp , October 23. [4] Ö. Yilmaz ad S. Rickard, Blid separatio of speech mixtures via time-frequecy maskig, IEEE Tras. Sigal Processig, vol. 52, o. 7, pp , July 24. [5] S. Witer, H. Sawada, S. Araki, ad S. Makio, Overcomplete bss for covolutive mixtures based o hierarchical clusterig, i Proc. ICA 24, Graada, Spai, September , pp [6] D. L. Wag ad G. J. Brow, Separatio of speech from iterferig souds based o oscillatory correlatio, IEEE Tras. Neural Networks, vol. 1, o. 3, pp , May [7] A. S. Bregma, Auditory Scee Aalysis, MIT Press, 2 editio, 199. [8] A. Jourjie, S. Richard, ad Ö. Yilmaz, Blid separatio of disjoit orthogoal sigals: Demixig sources from 2 mixtures, i Proc. ICASSP 2, Istabul, Turkey, Jue 2, vol. V, pp [9] G. Hu ad D. L. Wag, Moaural speech segregatio based o pitch trackig ad amplitude modulatio, IEEE Tras. Neural Networks, vol. 15, o. 5, pp , September 24. [1] S. Araki, S. Makio, H. Sawada, ad R. Mukai, Reducig musical oise by a fie-shift overlap-add method applied to source separatio usig a time-frequecy mask, i Proc. ICASSP25, March , vol. III, pp [11] D. L. Wag, O ideal biary mask as the computatioal goal of auditory scee aalysis, i Speech Separatio by Humas ad Machies, Pierre Diveyi, Ed., pp Kluwer, Norwell, MA, 25. [12] S. Araki, S. Makio, H. Sawada, ad R. Mukai, Uderdetermied blid separatio of covolutive mixtures of speech with directivity patter based mask ad ica, i Proc. ICA 24, September , pp [13] D. Kolossa ad R. Orglmeister, Noliear postprocessig for blid speech separatio, i Proc. ICA 24, Graada, Spai, September , pp [14] M. Ito, Y. Takeuchi, T. Matsumoto, H. Kudo, M. Kawamoto, T. Mukai, ad N. Ohishi, Movig-source separatio usig directioal microphoes, i Proc. ISSPIT 22, December 22, pp [15] A. J. Bell ad T. J. Sejowski, A iformatiomaximizatio approach to blid separatio ad blid decovolutio, Neural Computatio, vol. 7, o. 6, pp , [16] H.B. Nielse, Ucmif - a algorithm for ucostraied, oliear optimizatio, Tech. Rep. IMM-TEC-19, IMM, Techical Uiversity of Demark, 21. [17] M. S. Bradstei ad D. B. Ward, Eds., Microphoe Arrays, Digital Sigal Processig. Spriger, 21. [18] M. Wax ad T. Kailath, Detectio of sigals by iformatio theoretic criteria, IEEE Tras. Acous., Speech ad Sigal Processig, vol. ASSP-33, o. 2, pp , April 1985.
Intelligent location of two simultaneously active acoustic emission sources: Part II
1 Itelliget locatio of two simultaeously active acoustic emissio sources: Part II Tadej Kosel ad Igor Grabec Faculty of Mechaical Egieerig, Uiversity of Ljubljaa, Aškerčeva 6, POB 394, SI-11 Ljubljaa,
More informationCHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER
95 CHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER 5.1 GENERAL Ru-legth codig is a lossless image compressio techique, which produces modest compressio ratios. Oe way of icreasig the compressio ratio of a ru-legth
More informationAPPLICATION NOTE UNDERSTANDING EFFECTIVE BITS
APPLICATION NOTE AN95091 INTRODUCTION UNDERSTANDING EFFECTIVE BITS Toy Girard, Sigatec, Desig ad Applicatios Egieer Oe criteria ofte used to evaluate a Aalog to Digital Coverter (ADC) or data acquisitio
More informationEncode Decode Sample Quantize [ ] [ ]
Referece Audio Sigal Processig I Shyh-Kag Jeg Departmet of Electrical Egieerig/ Graduate Istitute of Commuicatio Egieerig M. Bosi ad R. E. Goldberg, Itroductio to Digital Audio Codig ad Stadards, Kluwer
More informationA SELECTIVE POINTER FORWARDING STRATEGY FOR LOCATION TRACKING IN PERSONAL COMMUNICATION SYSTEMS
A SELETIVE POINTE FOWADING STATEGY FO LOATION TAKING IN PESONAL OUNIATION SYSTES Seo G. hag ad hae Y. Lee Departmet of Idustrial Egieerig, KAIST 373-, Kusug-Dog, Taejo, Korea, 305-70 cylee@heuristic.kaist.ac.kr
More informationFingerprint Classification Based on Directional Image Constructed Using Wavelet Transform Domains
7 Figerprit Classificatio Based o Directioal Image Costructed Usig Wavelet Trasform Domais Musa Mohd Mokji, Syed Abd. Rahma Syed Abu Bakar, Zuwairie Ibrahim 3 Departmet of Microelectroic ad Computer Egieerig
More informationA New Space-Repetition Code Based on One Bit Feedback Compared to Alamouti Space-Time Code
Proceedigs of the 4th WSEAS It. Coferece o Electromagetics, Wireless ad Optical Commuicatios, Veice, Italy, November 0-, 006 107 A New Space-Repetitio Code Based o Oe Bit Feedback Compared to Alamouti
More informationMeasurement of Equivalent Input Distortion AN 20
Measuremet of Equivalet Iput Distortio AN 2 Applicatio Note to the R&D SYSTEM Traditioal measuremets of harmoic distortio performed o loudspeakers reveal ot oly the symptoms of the oliearities but also
More informationRadar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1
Advaces i Egieerig Research, volume 8 d Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 7) Radar emitter recogitio method based o AdaBoost ad decisio tree Tag Xiaojig,
More informationCombined Scheme for Fast PN Code Acquisition
13 th Iteratioal Coferece o AEROSPACE SCIENCES & AVIATION TECHNOLOGY, ASAT- 13, May 6 8, 009, E-Mail: asat@mtc.edu.eg Military Techical College, Kobry Elkobbah, Cairo, Egypt Tel : +(0) 4059 4036138, Fax:
More informationPROJECT #2 GENERIC ROBOT SIMULATOR
Uiversity of Missouri-Columbia Departmet of Electrical ad Computer Egieerig ECE 7330 Itroductio to Mechatroics ad Robotic Visio Fall, 2010 PROJECT #2 GENERIC ROBOT SIMULATOR Luis Alberto Rivera Estrada
More informationPermutation Enumeration
RMT 2012 Power Roud Rubric February 18, 2012 Permutatio Eumeratio 1 (a List all permutatios of {1, 2, 3} (b Give a expressio for the umber of permutatios of {1, 2, 3,, } i terms of Compute the umber for
More information4. INTERSYMBOL INTERFERENCE
DATA COMMUNICATIONS 59 4. INTERSYMBOL INTERFERENCE 4.1 OBJECT The effects of restricted badwidth i basebad data trasmissio will be studied. Measuremets relative to itersymbol iterferece, usig the eye patter
More informationSimulation of Microphone Array for Sound Localization using Human Binaural Hearing Model
Simulatio of Microphoe Array for Soud Localizatio usig Huma Biaural Hearig Model Viktor Hristov, Sejaa Pleshkova Abstract Ivestigatio of the huma hearig system ad attempts for its simulatio are oe the
More informationProblem of calculating time delay between pulse arrivals
America Joural of Egieerig Research (AJER) 5 America Joural of Egieerig Research (AJER) e-issn: 3-847 p-issn : 3-936 Volume-4, Issue-4, pp-3-4 www.ajer.org Research Paper Problem of calculatig time delay
More informationA New Design of Log-Periodic Dipole Array (LPDA) Antenna
Joural of Commuicatio Egieerig, Vol., No., Ja.-Jue 0 67 A New Desig of Log-Periodic Dipole Array (LPDA) Atea Javad Ghalibafa, Seyed Mohammad Hashemi, ad Seyed Hassa Sedighy Departmet of Electrical Egieerig,
More informationAn Adaptive Image Denoising Method based on Thresholding
A Adaptive Image Deoisig Method based o Thresholdig HARI OM AND MANTOSH BISWAS Departmet of Computer Sciece & Egieerig Idia School of Mies, Dhabad Jharkad-86004 INDIA {hariom4idia, matoshb}@gmail.com Abstract
More informationSummary of Random Variable Concepts April 19, 2000
Summary of Radom Variable Cocepts April 9, 2000 his is a list of importat cocepts we have covered, rather tha a review that derives or explais them. he first ad primary viewpoit: A radom process is a idexed
More informationx y z HD(x, y) + HD(y, z) HD(x, z)
Massachusetts Istitute of Techology Departmet of Electrical Egieerig ad Computer Sciece 6.02 Solutios to Chapter 5 Updated: February 16, 2012 Please sed iformatio about errors or omissios to hari; questios
More informationIntermediate Information Structures
Modified from Maria s lectures CPSC 335 Itermediate Iformatio Structures LECTURE 11 Compressio ad Huffma Codig Jo Roke Computer Sciece Uiversity of Calgary Caada Lecture Overview Codes ad Optimal Codes
More informationSingle Bit DACs in a Nutshell. Part I DAC Basics
Sigle Bit DACs i a Nutshell Part I DAC Basics By Dave Va Ess, Pricipal Applicatio Egieer, Cypress Semicoductor May embedded applicatios require geeratig aalog outputs uder digital cotrol. It may be a DC
More informationCancellation of Multiuser Interference due to Carrier Frequency Offsets in Uplink OFDMA
Cacellatio of Multiuser Iterferece due to Carrier Frequecy Offsets i Upli OFDMA S. Maohar, V. Tiiya, D. Sreedhar, ad A. Chocaligam Departmet of ECE, Idia Istitute of Sciece, Bagalore 56001, INDIA Abstract
More informationSpread Spectrum Signal for Digital Communications
Wireless Iformatio Trasmissio System Lab. Spread Spectrum Sigal for Digital Commuicatios Istitute of Commuicatios Egieerig Natioal Su Yat-se Uiversity Spread Spectrum Commuicatios Defiitio: The trasmitted
More informationEMPIRICAL MODE DECOMPOSITION IN AUDIO WATERMARKING BY USING WAVELET METHOD
EMPIRICAL MODE DECOMPOSITION IN AUDIO WATERMARKING BY USING WAVELET METHOD G.PAVAN KUMAR (PG Scholor) S.RAHUL (M.Tech) Departmet of ECE, S.V.College of Egieerig ad Techology, JNTUH Assistat Professor,
More information信號與系統 Signals and Systems
Sprig 24 信號與系統 Sigals ad Systems Chapter SS- Sigals ad Systems Feg-Li Lia NTU-EE Feb4 Ju4 Figures ad images used i these lecture otes are adopted from Sigals & Systems by Ala V. Oppeheim ad Ala S. Willsky,
More information信號與系統 Signals and Systems
Sprig 2 信號與系統 Sigals ad Systems Chapter SS- Sigals ad Systems Feg-Li Lia NTU-EE Feb Ju Figures ad images used i these lecture otes are adopted from Sigals & Systems by Ala V. Oppeheim ad Ala S. Willsky,
More informationDesign of FPGA- Based SPWM Single Phase Full-Bridge Inverter
Desig of FPGA- Based SPWM Sigle Phase Full-Bridge Iverter Afarulrazi Abu Bakar 1, *,Md Zarafi Ahmad 1 ad Farrah Salwai Abdullah 1 1 Faculty of Electrical ad Electroic Egieerig, UTHM *Email:afarul@uthm.edu.my
More informationNonlinear System Identification Based on Reduced Complexity Volterra Models Guodong Jin1,a* and Libin Lu1,b
6th Iteratioal Coferece o Electroics, Mechaics, Culture ad Medicie (EMCM 205) Noliear System Idetificatio Based o Reduced Complexity Volterra Models Guodog Ji,a* ad Libi Lu,b Xi a research istitute of
More informationCOMPRESSION OF TRANSMULTIPLEXED ACOUSTIC SIGNALS
COMPRESSION OF TRANSMULTIPLEXED ACOUSTIC SIGNALS Mariusz Ziółko, Przemysław Sypka ad Bartosz Ziółko Departmet of Electroics, AGH Uiversity of Sciece ad Techology, al. Mickiewicza 3, 3-59 Kraków, Polad,
More informationComparison of Frequency Offset Estimation Methods for OFDM Burst Transmission in the Selective Fading Channels
Compariso of Frequecy Offset Estimatio Methods for OFDM Burst Trasmissio i the Selective Fadig Chaels Zbigiew Długaszewski Istitute of Electroics ad Telecommuicatios Pozań Uiversity of Techology 60-965
More informationAnalysis of SDR GNSS Using MATLAB
Iteratioal Joural of Computer Techology ad Electroics Egieerig (IJCTEE) Volume 5, Issue 3, Jue 2015 Aalysis of SDR GNSS Usig MATLAB Abstract This paper explais a software defied radio global avigatio satellite
More informationUnit 5: Estimating with Confidence
Uit 5: Estimatig with Cofidece Sectio 8.2 The Practice of Statistics, 4 th editio For AP* STARNES, YATES, MOORE Uit 5 Estimatig with Cofidece 8.1 8.2 8.3 Cofidece Itervals: The Basics Estimatig a Populatio
More informationCompound Controller for DC Motor Servo System Based on Inner-Loop Extended State Observer
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 6, No 5 Special Issue o Applicatio of Advaced Computig ad Simulatio i Iformatio Systems Sofia 06 Prit ISSN: 3-970; Olie ISSN:
More informationThe Detection of Abrupt Changes in Fatigue Data by Using Cumulative Sum (CUSUM) Method
Proceedigs of the th WSEAS Iteratioal Coferece o APPLIED ad THEORETICAL MECHANICS (MECHANICS '8) The Detectio of Abrupt Chages i Fatigue Data by Usig Cumulative Sum (CUSUM) Method Z. M. NOPIAH, M.N.BAHARIN,
More informationApplication of Improved Genetic Algorithm to Two-side Assembly Line Balancing
206 3 rd Iteratioal Coferece o Mechaical, Idustrial, ad Maufacturig Egieerig (MIME 206) ISBN: 978--60595-33-7 Applicatio of Improved Geetic Algorithm to Two-side Assembly Lie Balacig Ximi Zhag, Qia Wag,
More informationINCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION
XIX IMEKO World Cogress Fudametal ad Applied Metrology September 6, 9, Lisbo, Portugal INCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION Dalibor
More informationUsing Color Histograms to Recognize People in Real Time Visual Surveillance
Usig Color Histograms to Recogize People i Real Time Visual Surveillace DANIEL WOJTASZEK, ROBERT LAGANIERE S.I.T.E. Uiversity of Ottawa, Ottawa, Otario CANADA daielw@site.uottawa.ca, lagaier@site.uottawa.ca
More informationWAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI
WAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI Muhammad Kabir McGill Uiversity Departmet of Electrical ad Computer Egieerig Motreal, QC H3A 2A7 Email: muhammad.kabir@mail.mcgill.ca Carlos Christofferse
More informationLaboratory Exercise 3: Dynamic System Response Laboratory Handout AME 250: Fundamentals of Measurements and Data Analysis
Laboratory Exercise 3: Dyamic System Respose Laboratory Hadout AME 50: Fudametals of Measuremets ad Data Aalysis Prepared by: Matthew Beigto Date exercises to be performed: Deliverables: Part I 1) Usig
More informationFault Diagnosis in Rolling Element Bearing Using Filtered Vibration and Acoustic Signal
Volume 8 o. 8 208, 95-02 ISS: 3-8080 (prited versio); ISS: 34-3395 (o-lie versio) url: http://www.ijpam.eu ijpam.eu Fault Diagosis i Rollig Elemet Usig Filtered Vibratio ad Acoustic Sigal Sudarsa Sahoo,
More informationMid-level representations for audio content analysis *Slides for this lecture were created by Anssi Klapuri
Mid-level represetatios for audio cotet aalysis *Slides for this lecture were created by Assi Klapuri Represetatios 1 Sources: Ellis, Rosethal, Mid-level represetatios for Computatioal Auditory Scee Aalysis,
More informationNovel pseudo random number generation using variant logic framework
Edith Cowa Uiversity Research Olie Iteratioal Cyber Resiliece coferece Cofereces, Symposia ad Campus Evets 011 Novel pseudo radom umber geeratio usig variat logic framework Jeffrey Zheg Yua Uiversity,
More informationRoberto s Notes on Infinite Series Chapter 1: Series Section 2. Infinite series
Roberto s Notes o Ifiite Series Chapter : Series Sectio Ifiite series What you eed to ow already: What sequeces are. Basic termiology ad otatio for sequeces. What you ca lear here: What a ifiite series
More informationLossless image compression Using Hashing (using collision resolution) Amritpal Singh 1 and Rachna rajpoot 2
Lossless image compressio Usig Hashig (usig collisio resolutio) Amritpal Sigh 1 ad Racha rajpoot 2 1 M.Tech.* CSE Departmet, 2 Departmet of iformatio techology Guru Kashi UiversityTalwadi Sabo, Bathida
More informationVoice Command Recognition System Based on MFCC and VQ Algorithms
Voice Commad Recogitio System Based o MFCC ad VQ Algorithms Mahdi Shaeh, ad Azizollah Taheri Abstract The goal of this project is to desig a system to recogitio voice commads. Most of voice recogitio systems
More informationOptimal Arrangement of Buoys Observable by Means of Radar
Optimal Arragemet of Buoys Observable by Meas of Radar TOMASZ PRACZYK Istitute of Naval Weapo ad Computer Sciece Polish Naval Academy Śmidowicza 69, 8-03 Gdyia POLAND t.praczy@amw.gdyia.pl Abstract: -
More informationDesign of FPGA Based SPWM Single Phase Inverter
Proceedigs of MUCEET2009 Malaysia Techical Uiversities Coferece o Egieerig ad Techology Jue 20-22, 2009, MS Garde,Kuata, Pahag, Malaysia MUCEET2009 Desig of FPGA Based SPWM Sigle Phase Iverter Afarulrazi
More informationLogarithms APPENDIX IV. 265 Appendix
APPENDIX IV Logarithms Sometimes, a umerical expressio may ivolve multiplicatio, divisio or ratioal powers of large umbers. For such calculatios, logarithms are very useful. They help us i makig difficult
More informationAC : USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM
AC 007-7: USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM Josue Njock-Libii, Idiaa Uiversity-Purdue Uiversity-Fort Waye Josué Njock Libii is Associate Professor
More informationNoncoherent Detection of Satellite-Based AIS Signal with Phase Offset Compensation
Sesors & Trasducers Vol. 166 Issue 3 March 2014 pp. 128-134 Sesors & Trasducers 2014 by IFSA Publishig S. L. http://www.sesorsportal.com Nocoheret Detectio of Satellite-Based AIS Sigal with Phase Offset
More informationOutline. Motivation. Analog Functional Testing in Mixed-Signal Systems. Motivation and Background. Built-In Self-Test Architecture
Aalog Fuctioal Testig i Mixed-Sigal s Jie Qi Dept. of Electrical & Computer Egieerig Aubur Uiversity Co-Advisors: Charles Stroud ad Foster Dai Outlie Motivatio ad Backgroud Built-I Self-Test Architecture
More informationA Data Field method for speech enhancement incorporating Binary Time-Frequency Masking
Jiaju HUANG Yafei ZHANG Xiogwei ZHANG Tao ZHU Istitute of Commad Automatio PLA Uiversity of Sciece ad Techology. A Data Field method for speech ehacemet icorporatig Biary Time-Frequecy Maskig Abstract.
More informationA Reduced Complexity Channel Estimation for OFDM Systems with Precoding and Transmit Diversity in Mobile Wireless Channels Hlaing Minn, Dong In Kim an
A Reduced Complexity Chael Estimatio for OFDM Systems with Precodig ad Trasmit Diversity i Mobile Wireless Chaels Hlaig Mi, Dog I Kim ad Vijay K. Bhargava Departmet of Electrical ad Computer Egieerig,
More informationLETTER A Novel Adaptive Channel Estimation Scheme for DS-CDMA
1274 LETTER A Novel Adaptive Chael Estimatio Scheme for DS-CDMA Che HE a), Member ad Xiao-xiag LI, Nomember SUMMARY This paper proposes a adaptive chael estimatio scheme, which uses differet movig average
More informationTotal Harmonics Distortion Reduction Using Adaptive, Weiner, and Kalman Filters
Wester Michiga Uiversity ScholarWorks at WMU Master's Theses Graduate College 6-2016 Total Harmoics Distortio Reductio Usig Adaptive, Weier, ad Kalma Filters Liqaa Alhafadhi Wester Michiga Uiversity, liquaa.alhafadhi@yahoo.com
More informationH2 Mathematics Pure Mathematics Section A Comprehensive Checklist of Concepts and Skills by Mr Wee Wen Shih. Visit: wenshih.wordpress.
H2 Mathematics Pure Mathematics Sectio A Comprehesive Checklist of Cocepts ad Skills by Mr Wee We Shih Visit: weshih.wordpress.com Updated: Ja 2010 Syllabus topic 1: Fuctios ad graphs 1.1 Checklist o Fuctios
More informationAdaptive Fuzzy Color Interpolation
Joural of Electroic Imagig Vol. (3), July 00. Adaptive Fuzzy Color Iterpolatio Pig-Sig Tsai, Tiku Acharya,, Ajay K. Ray 3 Itel Corporatio Desktop Architecture Lab 5000 West Chadler Boulevard Mail Stop:
More informationHigh Speed Area Efficient Modulo 2 1
High Speed Area Efficiet Modulo 2 1 1-Soali Sigh (PG Scholar VLSI, RKDF Ist Bhopal M.P) 2- Mr. Maish Trivedi (HOD EC Departmet, RKDF Ist Bhopal M.P) Adder Abstract Modular adder is oe of the key compoets
More informationA Study on Performance Analysis for Error Probability in SWSK Systems
556 Tae-Il Jeog, et al.: A STUDY ON PERFORMANCE ANALYSIS FOR ERROR PROBABILITY IN SWSK SYSTEMS A Study o Performace Aalysis for Error Probability i SWSK Systems Tae-Il Jeog, wag-seo Moo, ad Jog-Nam Kim,
More informationSELEX Elsag. 5/18/2012 R. Pucci SDR 12 WinnComm 1
SELEX Elsag 5/18/01 R. Pucci SDR 1 WiComm 1 Military BU - SELEX Elsag Possible update of SDR Platforms to COGNITIVE architectures COGNITIVE MANAGER INTERFACE Geolocatio, Voice, Video, etc Applicatio Policy
More informationSelection of the basic parameters of the lens for the optic-electronic target recognition system
Proceedigs of the 5th WSEAS It. Cof. o COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS AND CYBERNETICS, Veice, Italy, November 0-, 006 317 Selectio of the basic parameters of the les for the optic-electroic
More informationHarnessing oversampling in correlation-coded OTDR
Haressig oversamplig i correlatio-coded OTDR Ruoli Liao, 1 Mig Tag, 1,* Ca Zhao, 1 Hao Wu, 1 Sogia Fu, 1 Demig Liu, 1 ad Perry Pig Shum 1 Wuha Natioal Laboratory for Optoelectroics (WNLO) ad Natioal Egieerig
More informationObjectives. Some Basic Terms. Analog and Digital Signals. Analog-to-digital conversion. Parameters of ADC process: Related terms
Objectives. A brief review of some basic, related terms 2. Aalog to digital coversio 3. Amplitude resolutio 4. Temporal resolutio 5. Measuremet error Some Basic Terms Error differece betwee a computed
More informationSampling. Introduction to Digital Data Acquisition: Physical world is analog CSE/EE Digital systems need to
Itroductio to Digital Data Acuisitio: Samplig Physical world is aalog Digital systems eed to Measure aalog uatities Switch iputs, speech waveforms, etc Cotrol aalog systems Computer moitors, automotive
More informationProcedia - Social and Behavioral Sciences 128 ( 2014 ) EPC-TKS 2013
Available olie at www.sciecedirect.com ScieceDirect Procedia - Social ad Behavioral Scieces 18 ( 014 ) 399 405 EPC-TKS 013 Iductive derivatio of formulae by a computer Sava Grozdev a *, Veseli Nekov b
More informationLecture 14. Design of audio WM Model of digital audio CO ( WAV-format): - dependent samples with frequency 44,1кHz, - amplitude of samples is
Lecture 14. Desig of audio WM Model of digital audio CO WAV-format: - depedet samples wit frequecy 44,1кHz, - amplitude of samples is quatized o 2 16 =65536 levels 16 bits Remark 1. Tere exists a lossy
More informationIntegrated Detection Method of Underwater Acoustic Fuze Based on IEMD, VIFD and ED
Advaces i Applied Acoustics Vol. Iss., November Itegrated Detectio Method of Uderwater Acoustic Fuze Based o IEMD, VIFD ad ED Qiao Hu Sciece ad Techology o Uderwater Iformatio ad Cotrol Laboratory, Xi
More informationCHAPTER 8 JOINT PAPR REDUCTION AND ICI CANCELLATION IN OFDM SYSTEMS
CHAPTER 8 JOIT PAPR REDUCTIO AD ICI CACELLATIO I OFDM SYSTEMS Itercarrier Iterferece (ICI) is aother major issue i implemetig a OFDM system. As discussed i chapter 3, the OFDM subcarriers are arrowbad
More informationDetection of unreliable measurements in multi-sensor devices
MeMeA 8 IEEE Iteratioal Workshop o Medical Measuremets ad Applicatios Ottawa, Otario, Caada - May 9-1, 8 Detectio of ureliable measuremets i multi-sesor devices Yedek Asfaw 1, Ady Adler yasfaw@site.uottawa.ca
More informationPerformance Analysis of Channel Switching with Various Bandwidths in Cognitive Radio
Performace Aalysis of Chael Switchig with Various Badwidths i Cogitive Radio Po-Hao Chag, Keg-Fu Chag, Yu-Che Che, ad Li-Kai Ye Departmet of Electrical Egieerig, Natioal Dog Hwa Uiversity, 1,Sec.2, Da-Hsueh
More informationMassachusetts Institute of Technology Dept. of Electrical Engineering and Computer Science Fall Semester, Introduction to EECS 2.
Massachusetts Istitute of Techology Dept. of Electrical Egieerig ad Computer Sciece Fall Semester, 006 6.08 Itroductio to EECS Prelab Exercises Pre-Lab#3 Modulatio, demodulatio, ad filterig are itegral
More informationCross-Entropy-Based Sign-Selection Algorithms for Peak-to-Average Power Ratio Reduction of OFDM Systems
4990 IEEE TRASACTIOS O SIGAL PROCESSIG, VOL. 56, O. 10, OCTOBER 2008 Cross-Etropy-Based Sig-Selectio Algorithms for Peak-to-Average Power Ratio Reductio of OFDM Systems Luqig Wag ad Chitha Tellambura Abstract
More informationTehrani N Journal of Scientific and Engineering Research, 2018, 5(7):1-7
Available olie www.jsaer.com, 2018, 5(7):1-7 Research Article ISSN: 2394-2630 CODEN(USA): JSERBR 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
More informationWSN Node Localization Regularization Algorithm Based on Quasi Optimal Criterion Parameter Selection
Sesors & rasducers Vol. 23 Special Issue July 203 pp. 94-98 Sesors & rasducers 203 by IFSA http://www.sesorsportal.com WSN Node Localizatio Regularizatio Algorithm Based o Quasi Optimal Criterio Parameter
More informationPERFORMANCE OF MULTIBEAM MIMO FOR NLOS MILLIMETER WAVE INDOOR COMMUNICATION SYSTEMS CHAPTER 6
PERFORMANCE OF MULTIBEAM MIMO FOR NLOS MILLIMETER WAVE INDOOR COMMUNICATION SYSTEMS CAPTER 6 1 PERFORMANCE OF MULTIBEAM MIMO FOR NLOS MILLIMETER WAVE INDOOR COMMUNICATION SYSTEMS CAPTER 6 igh data rate
More informationAdaptive Resource Allocation in Multiuser OFDM Systems
Adaptive Resource Allocatio i Multiuser OFDM Systems Fial Report Multidimesioal Digital Sigal Processig Malik Meherali Saleh The Uiversity of Texas at Austi malikmsaleh@mail.utexas.edu Sprig 005 Abstract
More informationAkinwaJe, A.T., IbharaJu, F.T. and Arogundade, 0.1'. Department of Computer Sciences University of Agriculture, Abeokuta, Nigeria
COMPARATIVE ANALYSIS OF ARTIFICIAL NEURAL NETWORK'S BACK PROPAGATION ALGORITHM TO STATISTICAL LEAST SQURE METHOD IN SECURITY PREDICTION USING NIGERIAN STOCK EXCHANGE MARKET AkiwaJe, A.T., IbharaJu, F.T.
More informationThe Firing Dispersion of Bullet Test Sample Analysis
Iteratioal Joural of Materials, Mechaics ad Maufacturig, Vol., No., Ma 5 The Firig Dispersio of Bullet Test Sample Aalsis Youliag Xu, Jubi Zhag, Li Ma, ad Yoghai Sha Udisputed, this approach does reduce
More information8. Combinatorial Structures
Virtual Laboratories > 0. Foudatios > 1 2 3 4 5 6 7 8 9 8. Combiatorial Structures The purpose of this sectio is to study several combiatorial structures that are of basic importace i probability. Permutatios
More informationCh 9 Sequences, Series, and Probability
Ch 9 Sequeces, Series, ad Probability Have you ever bee to a casio ad played blackjack? It is the oly game i the casio that you ca wi based o the Law of large umbers. I the early 1990s a group of math
More informationAnalysis and Optimization Design of Snubber Cricuit for Isolated DC-DC Converters in DC Power Grid
Aalysis ad Optimizatio Desig of Subber Cricuit for Isolated DC-DC Coverters i DC Power Grid Koji Orikawa Nagaoka Uiversity of Techology Nagaoka, Japa orikawa@st.agaokaut.ac.jp Ju-ichi Itoh Nagaoka Uiversity
More informationEstimation of non Distortion Audio Signal Compression
Estimatio of o Distortio Audio Sigal Compressio M. BAK, S. PODOXI, V. TSIGOUZ Departmet of Commuicatio Egieerig Ceter for Techological Educatio Holo 5 Golomb street, POB 35 Holo 58, Tel: +97-3-5669 ISRAEL
More informationPulse-echo Ultrasonic NDE of Adhesive Bonds in Automotive Assembly
ECNDT 6 - Poster 7 Pulse-echo Ultrasoic NDE of Adhesive Bods i Automotive Assembly Roma Gr. MAEV, Sergey TITOV, Uiversity of Widsor, Widsor, Caada Abstract. Recetly, adhesive bodig techology has begu to
More informationDiscrete Mathematics and Probability Theory Spring 2014 Anant Sahai Note 12
EECS 70 Discrete Mathematics ad Probability Theory Sprig 204 Aat Sahai Note 2 Probability Examples Based o Coutig We will ow look at examples of radom experimets ad their correspodig sample spaces, alog
More informationA Novel Three Value Logic for Computing Purposes
Iteratioal Joural o Iormatio ad Electroics Egieerig, Vol. 3, No. 4, July 23 A Novel Three Value Logic or Computig Purposes Ali Soltai ad Saeed Mohammadi Abstract The aim o this article is to suggest a
More informationCCD Image Processing: Issues & Solutions
CCD Image Processig: Issues & Solutios Correctio of Raw Image with Bias, Dark, Flat Images Raw File r x, y [ ] Dark Frame d[ x, y] Flat Field Image f [ xy, ] r[ x, y] d[ x, y] Raw Dark f [ xy, ] bxy [,
More informationApplication of Image Fusion to Wireless Image Transmission
Applicatio of Image Fusio to Wireless Image Trasmissio Liae C. Ramac ad Pramod K. Varshey EECS Departmet, Syracuse Uiversity 121 Lik Hall, Syracuse, NY 13244 Email: lcramac@mailbox.syr.edu, varshey@cat.syr.edu
More informationVIBRATION FEATURE EXTRACTION METHODS FOR GEAR FAULTS DIAGNOSIS -A REVIEW
VIBRATION FEATURE EXTRACTION METHODS FOR GEAR FAULTS DIAGNOSIS -A REVIEW Nioslav Zuber, Rusmir Bajrić, Draga Cvetković 3 Faculty of Techical Scieces, Uiversity of Novi Sad, Trg Dositeja Obradovića 6, Serbia
More informationAN ESTIMATION OF MULTILEVEL INVERTER FED INDUCTION MOTOR DRIVE
9 IJRIC. All rights reserved. IJRIC www.ijric.org E-ISSN: 76-3336 AN ESTIMATION OF MULTILEVEL INVERTER FED INDUCTION MOTOR DRIVE K.RAMANI AND DR.A. KRISHNAN SMIEEE Seior Lecturer i the Departmet of EEE
More informationSimulation and Analysis on Signal Acquisition of BDS Receiver with The Aid of INS
Modelig Simulatio ad Optimizatio Techologies ad Applicatios (MSOTA 2016 Simulatio ad Aalysis o Sigal Acquisitio of DS Receiver with The Aid of INS Shulei Che Xiaqig Tag Xuwei Cheg Juqiag Gao ad Zepeg Su
More informationJoint Power Allocation and Beamforming for Cooperative Networks
It. J. Commuicatios, etwork ad System Scieces,, 4, 447-45 doi:.436/ijcs..4753 Published Olie July (http://www.scirp.org/joural/ijcs) Joit Power Allocatio ad Beamformig for Cooperative etworks Sodes Maadi,,
More informationTHE OCCURRENCE OF TRANSIENT FIELDS AND ESD IN TYPICAL SELECTED AREAS
THE OCCURRENCE OF TRANSIENT FIELDS AND ESD IN TYPICAL SELECTED AREAS Stepha FREI Techical Uiversity Berli, Istitute of Electrical Power Egieerig Eisteiufer, 587 Berli, Germay e-mail: frei@ihs.ee.tu-berli.de
More informationRobust Audio Watermarking Using Multiwavelet Transform and Genetic Algorithm
Robust Audio Watermarkig Usig Multiwavelet Trasform ad Geetic Algorithm PRAYOTH KUMSAWAT 1, KITTI ATTAKITMONGCOL 2 AND ARTHIT SRIKAEW 2 1 School of Telecommuicatio Egieerig, 2 School of Electrical Egieerig
More informationlecture notes September 2, Sequential Choice
18.310 lecture otes September 2, 2013 Sequetial Choice Lecturer: Michel Goemas 1 A game Cosider the followig game. I have 100 blak cards. I write dow 100 differet umbers o the cards; I ca choose ay umbers
More informationThe Institute of Chartered Accountants of Sri Lanka
The Istitute of Chartered Accoutats of Sri Laka Postgraduate Diploma i Busiess ad Fiace Quatitative Techiques for Busiess Hadout 02:Presetatio ad Aalysis of data Presetatio of Data The Stem ad Leaf Display
More informationBy: Pinank Shah. Date : 03/22/2006
By: Piak Shah Date : 03/22/2006 What is Strai? What is Strai Gauge? Operatio of Strai Gauge Grid Patters Strai Gauge Istallatio Wheatstoe bridge Istrumetatio Amplifier Embedded system ad Strai Gauge Strai
More information3. Error Correcting Codes
3. Error Correctig Codes Refereces V. Bhargava, Forward Error Correctio Schemes for Digital Commuicatios, IEEE Commuicatios Magazie, Vol 21 No1 11 19, Jauary 1983 Mischa Schwartz, Iformatio Trasmissio
More informationAssessment of Soil Parameter Estimation Errors for Fusion of Multichannel Radar Measurements
Assessmet of Soil Parameter Estimatio Errors for Fusio of Multichael Radar Measuremets A. urei, D. Marshall, D. Radford,. Lever Cardiff Uiversity 5 The Parade, Roath, CF4 0YF, Cardiff, U A.urei@cs.cf.ac.u
More informationImplementation of OFDM wireless communication model for achieving the improved BER using DWT-OFDM
www.ijecs.i Iteratioal Joural Of Egieerig Ad Computer Sciece ISSN: 2319-7242 Volume 6 Issue 1 Ja. 2017, Page No. 19951-19959 Idex Copericus Value (2015): 58.10, DOI: 10.18535/ijecs/v6i1.17 Implemetatio
More informationECE 333: Introduction to Communication Networks Fall Lecture 4: Physical layer II
ECE 333: Itroductio to Commuicatio Networks Fall 22 Lecture : Physical layer II Impairmets - distortio, oise Fudametal limits Examples Notes: his lecture cotiues the discussio of the physical layer. Recall,
More information