基于相关矩阵行列式分析的GSC二元麦克风阵列语音增强算法

Dual microphone GSC speech enhancement based on determinant analysis of correlation matrix

为了提高广义旁瓣相消器语音增强算法在二元麦克风阵列中的噪声抑制能力,提出了一种基于相关矩阵行列式分析的GSC二元麦克风阵列语音增强算法,该方法首先对二元阵列的输入进行相关矩阵行列式分析,利用基于语音活动检测和信噪比估计确定系数的维纳滤波器改进GSC结构的固定波束成形支路的输出,再利用基于相关矩阵行列式分析的语音活动检测更新自适应噪声抵消器系数,提高GSC结构中自适应支路噪声估计的准确性.实验结果表明,相较于近些年来其他二元麦克风阵列的语音增强方法,该方法在多个噪声源和复杂类型噪声条件下的语音质量更高,信号失真更小,处理后的语音更接近目标语音.

To improve the denoising ability of the generalized sidelobe canceller algorithm for dual small microphone,a new GSC algorithm based on determinant analysis of correlation matrix is proposed.In the fixed beam forming branch of GSC,the method first analyzes the determinant of the correlation matrix of the input,then uses voice activity detection and SNR estimation to determine wiener filter coefficients,uses wiener filter to improve the output of fixed beam former.In adaptive beam forming branch of GSC,uses speech activity detection based on determinant analysis to update coefficients of adaptive noise canceller and improve the accuracy of noise estimation.The experimental results show that compared with other speech enhancement methods of dual small microphone array in recent years,this method has higher speech quality,less signal distortion and is closer to the target speech under the condition of multiple noise sources and complex types of noise.