Noise reduction for hearing aids using digital signal processing

1.降噪算法提出了一种降低平稳/非平稳环境噪声的算法，并通过计算机仿真进行了验证。此外，还利用信号处理系统构建了总体降噪系统。在环境噪声抑制方面，将自适应波束形成技术与传统的固定波束形成器相结合，不仅对平稳噪声，而且对非平稳噪声都能获得稳健的衰减。另一方面，为了减少咆哮，引入了自适应反馈抵消器。评价该系统已应用于实际环境噪声中，并从降噪量等方面进行了评价。结果表明：对于环境噪声，功率较大的方向性噪声降低了20分贝以上，混响等非方向性噪声也降低了10分贝；对于咆哮声，咆哮幅度增加了20分贝。将改进的语音清晰度降噪系统应用于被噪声污染的语音，并进行了听力测试。主要研究结果如下：在环境降噪方面，通过引入该系统，语音清晰度从50%提高到70%。对于咆哮对消器，传统的自适应反馈对消器会对诸如浊音等周期性信号产生干扰。然而，在我们的系统中，频率压缩和自适应反馈对消器的结合使用产生的干扰很小，并且对语音清晰度没有影响。

1. Algorithm for noise reduction An algorithm for reducing stationary/nonstationary environmental noise has been developed and has been verified via computer simulation. Moreover, total noise reduction system has been constructed using signal processing system. For ebvironmental noise reduction, an adaptive beamforming technique was combined with conventional fixed beamformer to obtain robust attenuation for not only stationary noise but also nonstationary noise. On the other hand, for reducing howling, adaptive feedback canceler was introduced.2. Evaluation The system developed has been applied to real environmental noise and was evaluated in terms of the amount of noise reduction and other issues. The results can be summarized as follows : As for the environmental noise, directional noise with large power was reduced by more than 20 dB,while nondirectional noise such as reverberation is also reduced by 10 dB.As for the howling, howling margin increased by 20 dB.3. Improvement in Speech Intelligibility Noise reduction system has been applied to the speech corrupted by noise, and the listening test was conducted. The results can be summarized as follows : As for the environmental noise reduction, the speech intelligibility was improved from 50 % to 70 % by introducing the proposed system. As for the howling cancellation, periodic signals such as voiced speech is corrupted by the conventional adaptive feedback canceler. However, use of frequency compressor together with adaptive feedback canceler as proposed in our system yields little corruption and no effect on speech intelligibility.