Advancements in encoded speech data by background noise suppression under uncontrolled environment

The enhancements made to an earlier proposed combination of spectral subtraction using voice activity detection (SS-VAD) and linear predictive coding (LPC) system under uncontrolled environments are demonstrated in this work. In the earlier proposed system, musical noise had an adverse effect on encoding performance. In the enhanced version, a minimum mean square error-spectrum power estimator based on the zero-crossing (MMSE-SPZC) enhancement block is introduced before the LPC. Objective and subjective performance evaluations are conducted for noisy and enhanced speech data. Through experiments based on the NOIZEUS speech data, we show that the proposed framework works significantly better than the earlier reported SS-VAD and LPC-based system. The source code of proposed algorithm is made publicly available at: https://sites.google.com/view/thimmarajayadavag/downloads