基于块对角结构的语音信号盲压缩重构

为在复杂环境中提高语音信号的重构精度, 提出基于正交块对角结构的语音信号盲压缩重构算法, 通过最优求解找到一组对应的正交块对角变换基及稀疏矩阵, 利用二者乘积实现语音信号压缩重构。该算法确保盲压缩感知理论具有唯一解, 能从复杂的环境中恢复原始语音信号, 具有更强的自适应性, 在保证听觉效果的同时, 大大降低了观测维度。实验结果表明, 基于正交块对角结构盲压缩感知(OBD-BCS: Orthonormal Block Diagonal Blind Compressed Sensing)算法能高质量恢复语音信号原始结构。

Blind Compressed Sensing Reconstruction of Speech Signal Based on Block Diagonal Structure

In order to improve the reconstruction accuracy of the speech signal in the complex environment, the reconstruction method based on orthonormal block diagonal blind compressed sensing theory is proposed, the method finds a pair of matrices by optimal solving, to reconstruct the speech signal. The algorithm takes on higher adaptability; ensures blind compressed sensing theory having a unique solution and can restore the original speech signal in a complex environment. The method ensures sound effects, and greatly reduces the observation dimension. Experimental result shows that blind the compressed sensing based on OBD-BCS (Orthogonal Block Diagonal Blind Compressed Sensing) structure algorithm can achieve high-quality reconstruction of a speech signal.