论文部分内容阅读
压缩感知(Compressed Sensing,CS)理论突破了经典采样定理的理论边界,为信号压缩提供了另一种途径。基于CS理论框架,做了两方面工作:为提高语音字典对信号的匹配性,设计了一种基于K-L展开的非相干语音字典;针对现有匹配追踪(MP,OMP)算法的不足,提出分段匹配追踪(Segment MP,SegMP)算法。首先对语音自相关函数进行建模并估计模型参数,构造语音自适应非相干字典,然后采用SegMP对语音稀疏向量分段观测,获得多个低维矢量,最后结合模型参数重建字典并重构信号,实现了语音压缩感知。语音测试结果表明:相比现有方案,本文方案对信号的稀疏表示更为精准,具有更好的重构质量,且降低了计算复杂度。
Compressed Sensing (CS) theory breaks through the theoretical boundary of the classical sampling theorem, providing another approach for signal compression. Based on the theoretical framework of CS, this paper has done two tasks: In order to improve the matching of speech dictionary to signal, a non-coherent speech dictionary based on KL expansion was designed. In view of the shortcomings of existing MP (OMP) algorithm, Segment MP (SegMP) algorithm. Firstly, the speech autocorrelation function is modeled and the model parameters are estimated. The speech adaptive incoherent dictionary is constructed. Segmented speech sparse vectors are segmented to obtain multiple low-dimensional vectors. Finally, the parameters are reconstructed and reconstructed , To achieve the voice compression perception. The result of speech test shows that compared with the existing schemes, the scheme proposed in this paper is more accurate for signal sparse representation, has better reconstruction quality and reduces computational complexity.