Abstract: The commonly used multi-channel loudness compensation algorithm in digital aids is to compensate loudness in each divided frequency band. This can have negative effects to the formant structure and lead to poor understanding of semantic information. A new method of nonlinear loudness compensation has been put forward. It does not need to divide the frequency band and can compensate the speech signal according to the hearing-impaired listeners’ realtime awareness to sound intensity on different frequency levels. This method not only makes the loudness of voice completely mapped to the listener’s hearing dynamic range, but also makes the loudness curve almost consistent with the original speech signal in outline. During the experiment, the use of wavelet de-noising and frequency shift technique ensures the processing effect more excellent. The experimental result indicates that the new method can effectively improve the speech loudness and the speech recognition rate for patients, and has practical value.

Key words: hearing impairment, digital aids, multi-channel loudness compensation, nonlinear loudness compensation

摘要： 数字助听器中的多通道响度补偿算法，通常是在划分出的多个频段内分别进行补偿，这在一定程度上会导致共振峰结构的破坏，造成语义信息的曲解或丢失。一种新的非线性的响度补偿方法的提出，无需进行频段的划分，并且根据听损患者在不同频率上对声音强弱的实时感知特性对语音信号进行补偿。该方法不仅使补偿后语音的响度完全映射到听者的听觉范围内，还使得补偿后的响度曲线在轮廓上与原始语音基本保持一致。实验中，小波去噪和移频技术的运用保证了处理效果更加优良。结果表明，该方法能有效提高语音响度和改善听损患者的言语识别率，具有一定的应用价值。

关键词: 听力损伤, 数字助听器, 多通道响度补偿, 非线性响度补偿