改进的变步长最小均方误差电子耳蜗语音增强算法

发布时间：2018-06-03 06:04

本文选题：电子耳蜗 + 噪声抑制　；参考：《计算机应用》2017年04期

【摘要】：针对外部强噪声环境下电子耳蜗语音质量受损、适应性差等问题,提出了基于谱减法和变步长最小均方误差(LMS)自适应滤波算法联合去噪的改进方法,并以该方法构建了一个电子耳蜗前端语音预处理系统。利用变步长LMS自适应滤波算法输出误差的平方项来调节步长,采用步长值固定与变化相结合的方法,解决了自适应滤波算法收敛速度慢、稳态误差大的问题,适应性得到提高,提高了语音信号通信质量。该系统以TMS320VC5416和音频编解码芯片TLV320AIC23B为核心,通过多通道缓冲串口(McBSP)和串行外设接口(SPI)实现了语音数据的高速采集和实时处理。实验仿真和测试结果表明该算法消除噪声性能好,信噪比在低输入信噪比情况下提高约10 d B,语音质量感知评价(PESQ)分值也得到较大提高,能有效提高语音信号质量,且该系统性能稳定,能进一步提高耳蜗前端语音的清晰度和可懂度。
[Abstract]:Aiming at the problems of damaged speech quality and poor adaptability of cochlear speech in the environment of strong external noise, an improved method based on spectrum subtraction and variable step size minimum mean square error (LMS) adaptive filtering algorithm is proposed. A speech preprocessing system for cochlear front end was constructed by this method. The variable step size LMS adaptive filtering algorithm is used to adjust the step size by using the square term of output error. The adaptive filtering algorithm solves the problem of slow convergence speed and large steady-state error, and improves the adaptability of the adaptive filtering algorithm. The communication quality of voice signal is improved. The system takes TMS320VC5416 and TLV320AIC23B as the core, and realizes the high-speed acquisition and real-time processing of voice data through multi-channel buffering serial port (McBSP) and serial peripheral interface (SPI). Experimental simulation and test results show that the proposed algorithm has good noise cancellation performance, the SNR is increased by about 10 dB under low input SNR, and the score of speech quality perception evaluation (PESQ) is also greatly improved, which can effectively improve the quality of speech signal. The system has stable performance and can further improve the articulation and intelligibility of cochlear front end speech.
【作者单位】：天津商业大学信息工程学院;
【基金】：国家自然科学基金资助项目(61401307) 天津市应用基础与前沿技术研究计划项目(14JCZDJC32600) 天津商业大学青年科研基金资助项目(150111) 国家级大学生创新创业训练计划项目(201610069085)~~
【分类号】：TN912.35

【相似文献】