Robust speech recognition techniques using a radial basis function neural network for mobile applications

We present the development of robust speech recognition techniques for voice activated dialing (VAD) for a cellular phone in a car. Noise reduction techniques are implemented before modeling the speech parameters in the homomorphic domain. The FFT derived cepstral coefficients are liftered and then Mel-scale warped to generate the feature vector. A radial basis function neural network is used as the final classifier and variation in its structure is studied to evaluate its real-time performance. Results on speaker-dependent and speaker-independent recognition across -5 dB to 25 dB SNR range are presented and a performance evaluation is carried out for different front-end techniques. The system is evaluated using the NOISEX-92 database