论文信息 - Signal Processing of Speech Using Independent Component Analysis Based on Information Maximization Algorithm

Signal Processing of Speech Using Independent Component Analysis Based on Information Maximization Algorithm

Signal Processing of Speech Using Independent Component Analysis Based on Information Maximization Algorithm Manabu KOTANI*, Satoshi MAEKAWA**, Seiichi OZAWA*** and Kenzo AKAZAWA* We describe what characteristics an independent component analysis can extract from Japanese continuous speech. Speech data was selected from ATR database uttered by a female speaker. The data was recorded at 20kHz sampling frequency and was pre-processed with a whitening filter. The learning algorithm of a network was an information-maximization approach proposed by Bell and Sejnowski. After the learning, most of the basis functions that are columns of a mixing matrix were localized in both time and frequency. Furthermore, we confirmed that there were some basis functions to extract the acoustic feature such as the pitch and the formant of each vowel.

[1] Terrence J. Sejnowski,et al. Learning Nonlinear Overcomplete Representations for Efficient Coding , 1997, NIPS.

[2] Christian Jutten,et al. Blind separation of sources, part I: An adaptive algorithm based on neuromimetic architecture , 1991, Signal Process..

[3] Terrence J. Sejnowski,et al. The “independent components” of natural scenes are edge filters , 1997, Vision Research.

[4] David J. Field,et al. Emergence of simple-cell receptive field properties by learning a sparse code for natural images , 1996, Nature.

[5] Terrence J. Sejnowski,et al. An Information-Maximization Approach to Blind Separation and Blind Deconvolution , 1995, Neural Computation.

[6] T J Sejnowski,et al. Learning the higher-order structure of a natural sound. , 1996, Network.