A Wavelet based Feature Extraction for Voice-Lock systems

In an approach to provide more secure, fast and efficient speaker verification system ("voice lock system"), this paper reviews the fast wavelet feature extraction technique and wavelet packet technique for speech recognition systems. One of the biggest advantages of the wavelet transform over the short time Fourier transform(STFT) is its capability to process non-stationary signals. As the prosodic features of speech signals in the voice lock systems play an important role, these features have to be extracted with great accuracy, which are efficiently extracted using the wavelet packets which provide the local time- spectral description. The adaptive frame synchronization method intelligently classifies these wavelet packets such that all the prosodic features are extracted efficiently. The paper also talks about the problems with the HMM systems in view of the security of the ASR systems.