An adaptive silence deletion algorithm for compression of telephone speech

The ever-increasing volume of speech generated for transmission and storage motivates a continuing search for an improved cost-effective means to effectively compress speech. Earlier work by various researchers has resulted in various compression algorithms. Of particular interest to us are those waveform compression techniques utilizing the principle of silence deletion developed by Gan (1988) and by Rose (1991) which was implemented on good quality speech signals. We further modify, optimize and evaluate these waveform compression algorithms, and test their viability and performance on noisy speech from actual telephone circuits. The results show that up to 50% compression of continuous speech is achievable at an acceptable level of playback quality.