摘要
实际麦克风阵列与假设阵列模型之间存在误差,导致麦克风阵列的语音定位和增强算法性能严重下降。该文提出了一种新的麦克风阵列误差有源校正方法。它假设麦克风阵列不存在位置误差,空间中放置一个位置精确已知的声源,通过对接收信号进行离散傅里叶变换,估计出各阵元的幅度,相位误差。该方法避开了估计协方差矩阵以及特征分解,以较小的运算量得到了较好的校正效果,适合工程应用。计算机仿真验证了该方法的有效性。
The errors between the actual microphone array and assumption array model exist. It results in heavily degrading of performance of speech localization and enhancement algorithms based on microphone. In this paper, a new active calibration approach for errors of microphone array is proposed. It assumes that microphone array dose not contain position errors and a sound source whose location is well known in the space. It can estimate the amplitude and phase errors through the discrete Fourier transformation of the received signals. This approach avoids to estimate the covariance matrix and eigen-decomposition. It has less computations with a good calibration performances and suitable for application in projects. Computer simulations verify its effectiveness.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2007年第S3期1505-1507,1511,共4页
Journal of University of Electronic Science and Technology of China
关键词
有源校正
麦克风阵列
模型误差
语音增强
语音定位
active calibration
microphone array
model errors
speech enhancement
speech localization
