Fu Q J, Shannon R V
Department of Auditory Implants and Perception, House Ear Institute, Los Angeles, California 90057, USA.
J Acoust Soc Am. 1998 Nov;104(5):2570-7. doi: 10.1121/1.423912.
It is widely assumed that the proper transformation of acoustic amplitude to electric amplitude is a critical factor affecting speech recognition in cochlear implant users and normal-hearing listeners. A four-channel noise-band speech processor was implemented, reducing spectral information to four bands. A power-law transformation was applied to the amplitude mapping stage in the speech processor design, and the exponent of the power function varied from a strongly compressive (p = 0.05) to a weakly compressive (p = 0.75) for implant listeners and from 0.3 to 3.0 for acoustic listeners. Results for implants showed that the best performance was achieved with an exponent of about 0.2, and performance gradually deteriorated when either more compressive or less compressive exponents were applied. The loudness growth functions of the four activated electrodes in each subject were measured and those data were well fit by a power function with a mean exponent of 2.72. The results indicated that the best performance was achieved when the normal loudness growth was restored. For acoustic listeners, results were similar to those observed with cochlear implant listeners, except that best performance was achieved with no amplitude nonlinearity (p = 1.0). The similarity of results in both acoustic and electric stimulation indicated that the performance deterioration observed for extreme nonlinearity was due to similar perceptual effects. The function relating amplitude mapping exponent and performance was relatively flat, indicating that phoneme recognition was only mildly affected by amplitude nonlinearity.
人们普遍认为，将声幅正确转换为电幅是影响人工耳蜗使用者和听力正常的聆听者语音识别的关键因素。实现了一种四通道噪声带语音处理器，将频谱信息减少到四个频段。在语音处理器设计的幅度映射阶段应用了幂律变换，对于植入式聆听者，幂函数的指数从强压缩（p = 0.05）到弱压缩（p = 0.75）变化，对于声学聆听者，指数从0.3到3.0变化。植入式的结果表明，指数约为0.2时性能最佳，而应用更强压缩或更弱压缩指数时性能会逐渐下降。测量了每个受试者中四个激活电极的响度增长函数，这些数据通过平均指数为2.72的幂函数得到了很好的拟合。结果表明，当恢复正常的响度增长时性能最佳。对于声学聆听者，结果与人工耳蜗聆听者的结果相似，只是在没有幅度非线性（p = 1.0）时性能最佳。声学刺激和电刺激结果的相似性表明，观察到的极端非线性导致的性能下降是由于相似的感知效应。幅度映射指数与性能之间的函数相对平缓，表明音素识别仅受到幅度非线性的轻微影响。
