多传声器小波多尺度信息融合滤波算法

为了滤掉信号采集过程中存在的环境噪声和信道噪声,提出一种基于小波多尺度信息融合和三角时延矢量误差的信号滤波算法。基于多传声器信号时延估计特性,提出时延矢量封闭准则。首先将多传声器同步采集所得信号进行小波多尺度分解,得到多尺度小波细节系数和近似系数,然后根据时延矢量封闭准则求取各层小波系数时延差,结合信息融合理论,提出一种三角时延矢量误差,由多传声器综合支持度获得时延阈值,将其作用于三角时延矢量误差,得到各层小波系数的权重,最后对小波系数重构得到降噪后的信号。实验表明,此滤波算法不仅滤掉了信号中的噪声成分,还优化了传声器资源的配置,既保留了高支持度传声器信号特性,还改善和提高了低支持度传声器信号质量,与小波分层阈值和全局阈值滤波算法相比充分体现出该算法的有效性。

Filtering algorithm for multi-microphones based on wavelet multi-scale information fusion

A signal filtering algorithm was proposed based on wavelet multi-scale information fusion and triangular time delay vector errors to filter the environment noises and channel noises in multi-microphone acquisition system.Time delay vector close rule is proposed based on the time delay characteristics of multi-microphone signals.The signals from multi-microphone system are decomposed with wavelet,and multi-scale wavelet coefficients are obtained.Time delay errors of multi-scale wavelet coefficients are calculated according to the time delay vector close rule.Triangular time delay vector errors are proposed based on information fusion and weighted with the time delay thresholds obtained from microphone confidences.Finally,the signals are reconstructed from the wavelet coefficients and corresponding weights.The proposed algorithm is compared with the global and layered threshold wavelet filtering algorithms,the results show that the proposed algorithm can effectively filter noises,optimize the configuration of the microphones,and has the advantages of information fusion.