www.bewet.fora.pl

loczytzn

,[link widoczny dla zalogowanych]
Based on multiple times and decorrelation convolution blind source separation

Below. Figure 3 shows the signal on the graph in Figure 1 above is the estimated source signal signal signal to the drawing of Figure 3 Figure 1 The figure is the estimate of the signal source signals. Defined as follows: SNRI = 101glO (such 2SNRIl + SNRI2SNRI =---- A (10) SNRI., SNRI:, SNR1 were the first and second separated signal to noise ratio to improve separation of signal rate and the average SNR increase rate. In the simulation, we obtained the average SNR was SNRI = 16.021. From the above two aspects can verify the validity of the algorithm. 5 Conclusion As the frequency domain, the interaction between each frequency is not correlation, convolution in the frequency domain to achieve blind signal separation, overcome the long-time domain filters in case of slow convergence and poor accuracy of the defect. convolution of time domain frequency domain corresponds to the product, which makes complex hybrid model in time domain convolution in the frequency domain is relatively simple, easy separation algorithm derivation, using the algorithm based on frequency domain complex environment are more likely to blind signal separation. signal propagation in the ocean, the sea surface and bottom of the signal reflection, but also by the shrimp, fish and various marine mammals due to noise pollution, making the acoustic signal separation very difficult, this article attempts to solutions based on multiple times and Convolution Blind Separation Algorithm related Record of water for the separation of acoustic signals, separation performance than other convolution in time domain separation algorithm has been improved significantly. References [I] PapaL,[link widoczny dla zalogowanych], SpenceC.Convolutiveblindseparationofnon-stationarysources [J]. IEEETransactionsonSpeechandAudioProcessing, 2000,8 (3) :320-327. [2] ParraL, SpenceC, DeVriesB.Convolutiveblindsoureeseparationbasedonmultipledeeorrelation [A]. NeuralNetworksforSignalProcessingProceedingsoftheIEEEWorkshop [C] ,1998,23-32. [3] WeinsteinE, FederM, Openheim.Multiehannelsig-nalseparationbydeeorrelation [J]. IEEETransactionsonSpeechandAudioProcessing, 1993, 1 (4) :405-413. [4] ParraL, SpeneeC.Separationofnon-stationarynaturalsig-nals [M]. IndependentComponentsAnalysis,[link widoczny dla zalogowanych], PrinciplesandPractice, CanbridgeUniversityPress ,2001.135-157. [5] IkramMZ, MorganDR.Exploringpermutationincon-sistencyinblindseparationofspeechsignalsinarever-berantenvironment �� A]. ICASSP.IEEEInternationalConferenceonAcoustics, SpeechandSignalProcessing [C] .2000.1041-1044. [6] IkramMZ,[link widoczny dla zalogowanych], MorganDR.Amuhiresolutionapproachtoblindseparationofspeechsiinareverberantenvironment [A]. ICASSP,[link widoczny dla zalogowanych], IEEElnterrmtionalConferenceonAcoustics, SpeechandSi ��1alProcessing [c] .2001.2757-2760.

相关的主题文章：

[link widoczny dla zalogowanych]

[link widoczny dla zalogowanych]

[link widoczny dla zalogowanych]