ROBUST AUTOMATIC SPEECH RECOGNITION IN LOW-SNR CAR ENVIRONMENTS BY THE APPLICATION OF A CONNECTIONIST SUBSPACE-BASED APPROACH TO THE MEL-BASED CEPSTRAL COEFFICIENTS

摘要

In this paper, the problem of robust large-vocabulary continuous-speech recognition (CSR) in the presence of highly interfering car noise has been considered. Our approach is based on the noise reduction of the parameters that we use for recognition, that is, the Mel-based cepstral coefficients. This is achieved by the use of a Multilayer Perceptron (MLP) network for noise reduction in the cepstral domain in order to get less-variant parameters. Then, the obtained enhanced features are refined via the Karhunen-Loeve Transform (KLT) implemented using the Principal Component Analysis (PCA). Experiments show that the use of the enhanced parameters using such an approach increases the recognition rate of the CSR process in highly interfering car noise environments. The HTK Hidden Markov Model Toolkit was used throughout our experiments. Results show that the proposed hybrid technique when included in the front-end of an HTK-based CSR system, outperforms that of the conventional recognition process based on either a KLT- or an MLP-based preprocessing recognition in severe interfering car noise environments for a wide range of SNRs varying from 16 dB to -4 dB using a noisy version of the TIMIT database.