Third International Conference on Spoken Language Processing (ICSLP 94)
This paper proposes a hearing aid which works by a new analysis-synthesis technique for telephonic speech. The software simulation will be introduced first and the hardware realization with multi-DSP units will be described. The developed system has two strong points in comparison with the other digital hearing aids. One is to use the special analysis-synthesis technique, and the other is in the amplitude compression method to reduce harmonic distortion which cannot be removed in instantaneous compression. Since the compression coefficients in this system change with rms level in a frame of signals, non-linear distortion little causes. The real-time system has been designed and produced with four DSP boards which have a quite same structure. As each DSP unit takes charge of independent program such as formant extraction, pitch extraction or amplitude compression etc., computer simulation programs can be transferred basically as they are. Assuming the auditory characteristics of sensorineural hearing loss, the effect of the processing by the DSP system has been investigated. The processing results show to be same as those by computer simulation. Furthermore, the preliminary listening tests indicate usefulness in this processing technique.
Bibliographic reference. Ikeda, Takashi / Tasaki, Kouji / Watanabe, Akira (1994): "A hearing aid by single resonant analysis for telephonic speech", In ICSLP-1994, 2003-2006.