A speech compression/expansion method based on subband filtering of the signal envelope
Aslak Bjerkvik, Asbjørn Krokstad, Barbara Resch · The Journal of the Acoustical Society of America · 2003
Compression algorithms are used in hearing aids for hearing-impaired listeners with recruitment, due to these listeners’ loss of dynamic range. We propose a method for both compressing and expanding speech signals with the goal to improve speech intelligibility in reverberant conditions where the important rapid variations of the signal are inherently reduced. The method is based on dividing the envelope of the speech signal into two subbands with a division around modulation frequencies of 2 Hz, and compressing the signal based on the lower envelope subband, while expanding the signal based on the higher envelope subband. The sub-band division is accomplished by filtering the envelope of the signal and also by computing separate envelopes for each subband. A secondary goal is to develop low-delay algorithms. Various methods for calculating the envelope of the signal are evaluated, with a focus on the delay. a)Currently at KTH, Speech processing group, Stockholm, Sweden.