Isolated vowel recognition using linear predictive features and neural network classifier fusion

J. Byorick1, R.P. Ramachandran1, R. Polikar1
1Department of Electrical and Computer Engineering, Rowan University, Glassboro, USA

Tóm tắt

In this work, various linear predictive feature vectors were used to train three different automated neural networks type classifiers for the task of isolated vowel recognition. The features used included linear prediction filter coefficients, reflection coefficients, log area ratios, and the linear predictive cepstrum. The three neural network classifiers used are the multilayer perceptron, radial basis function and the probabilistic neural network. The linear predictive cepstrum of dimension 12 is the best feature especially when training is done on clean speech and testing is done on noisy speech. Three different classifier fusion strategies (linear fusion, majority voting and weighted majority voting) were found to improve the performance. Linear fusion with varying weights is the best method and is most robust to noise.

Từ khóa

#Speech recognition #Neural networks #Cepstrum #Multi-layer neural network #Voting #Vectors #Nonlinear filters #Reflection #Multilayer perceptrons #Testing

Tài liệu tham khảo

littlestone, 1994, Weighted Majority Algorithm, 108, 212 10.1109/34.946993 ho, 1994, Decision combination in multiple classifier systems, IEEE Trans on Pattern Analysis and Machine Intelligence, 16, 66, 10.1109/34.273716 10.1109/ICASSP.2002.1004755 10.1109/NNSP.1999.788163 10.1109/21.155943 10.1016/S0167-6393(99)00077-1 10.1109/ICOSP.2000.891631 rabinerschafer, 1978, Digital Processing of Speech Signals 10.1109/ICASSP.1998.675468 duda, 2001, Pattern Classification 10.1109/79.180705 10.1109/MASSP.1987.1165576 10.1109/WISE.2000.882421 10.1016/S0031-3203(01)00235-7 haykins, 1999, Neural Networks A Comprehensive Foundation