An application of the merging algorithm with the discrete wavelet transform to extract valid speech-sound
Jin Ok Kim, Dae Jun Hwang, Han Wook Paek, Chin Hyun Chung · 2002
A valid speech-sound block can be classified to provide important information for speech recognition. The classification of the speech-sound block comes from the MRA (multi-resolution analysis) property of the DWT (discrete wavelet transform) which is used to reduce the computational time for the pre-processing of speech recognition. The merging algorithm is proposed to consider its position and frequency of the valid speech-sound blocks. The merging algorithm needs some numerical methods for an adaptive DWT implementation and shows the unvoiced/voiced classification and the denoising. So, it shows the improved performance especially in the signal added high frequency noise. The merging algorithm has the adaptive feature for arbitrary system noise and the excellent denoising SNR. Its extraction shows that the denoising of compounded noise and the improved extraction and the merging algorithm cannot be disturbed by an unexpected system interference.