【作者】 梁维谦； 郑方； 郑佳春； 朴志刚；

【Author】 LIANG Weiqian;ZHENG Fang;ZHENG Jiachun;PIAO Zhigang;Key Laboratory of Maritime Communication and Intelligent Electronic Systems of Fujian Province, College of Information Engineering, Jimei University;Center for Speech and Language Technologies, Research Institute of Information Technology, Tsinghua University;Xiamen LA and Associates Medical Equipment Co.LTD.;

【机构】 集美大学信息工程学院海上通信与智能电子系统福建省高等学校重点实验室； 清华大学信息技术研究院语音和语言技术中心； 厦门莱亚特医疗器械有限公司；

【摘要】 助听器对声音进行压缩放大,需求高言语清晰度的噪声算法。本文提出了一种子带自适应噪声抑制方法,通过加权重叠相加滤波器组和基于心理声学模型的子带划分、基于先验和后验信噪比的快变的非线性降噪增益、基于噪声声压级估值的慢变的增益下限阈值、基于峰值跟踪的子带增益平滑及其跟踪和释放时间系数的精细选择等算法,明显提高了言语清晰度。主观测听实验表明,本文方法对输入的不同信噪比的带噪语音的言语清晰度提高约12%到45%。我们在EZAIRO5900数字信号处理器上实现了本文方法,通过对增益公式的量化处理使得整个算法的运行效率提高约30%。更多还原

【Abstract】 In this paper, a sub-band based adaptive noise reduction algorithm was proposed to improve speech intelligibility in hearing aids. Weighted overlap-add filter bank and psycho-acoustic model were utilized to analyze and realize the sub-band splitting. The non-linear noise reduction gains were computed with the estimated a posteriori signal to noise ratio(SNR) and a priori SNR. The gain floors were determined with the estimated noise level in d B sound pressure level. The final gains were smoothed between the frames by the peak detector with fine selected attack and release time constants. Experimental results from listening tests showed that about 12% to 45% improvements in intelligibility can be achieved by applying the proposed algorithm to noise corrupted speech. A quantified gain table is also used to replace the non-linear gain computing when the algorithm was implemented on the EZAIRO5900 digital signal processor, and the execution cycle is reduced about30%.更多还原