Speaker Recognition Algorithm Based On ICA And ASR Speech Feature Selection

Speaker recognition,also known as voiceprint recognition,has been widely used in various security areas,Internet applications and communications areas.Usually,the process of speaker recognition involves extracting and identifying unique characteristics of the speech features from a group of speakers.For improving the performance of the speaker recognition system,it is important to select the efficient speech features.In this paper,we introduce Independent Component Analysis(ICA)and Analysis Sparse Representation(ASR)for speech features selection.1.The acoustic independent component(AICA)feature is proposed based on the ICA.Firstly,Gammatone Frequency coefficients of the different resolutions(DGFCs)are calculated from the speaker's speech signal.Then,the ICA speaker model is established used to selection AICA features from the DGFCs by the FastICA algorithm.Finally,SVM is used to realize the speaker recognition.The experiment results illustrate the effectiveness of the proposed system.2.The long-term acoustic(LTA)features is proposed for text-independent speaker recognition,which is a sparse presentation of the static features and dynamic information for speaker speech.First,the speech signal is segmented into frames which are overlapping with each other,and then the MFCCs frame features can be extracted to construct the super MFCCs frame by stacking some following frames of the current frame to capture the dynamic information of speech signal.The super MFCCs frames can be combined into a 2-D MFCCs features map(MFCCsmap).Finally,the speaker model can be built based on the analysis sparse model and the sparse representations of the MFCCsmap are used as the LTA features.The deep neural network(DNN)is employed as a classifier for speaker recognition.Experimental results illustrate the effectiveness of the proposed system.