Frame-based Adaptive Compressed Sensing of Speech Signal with LP Analysis
MA Yong-yang · Journal of Hangzhou Dianzi University · 2012
This paper focuses on the application of compressed sensing(CS) to nature speech signals.In linear prediction(LP) analysis,speech signals are regarded as sparse in the LP basis made by its own set of LP coefficients.Meanwhile,due to the short-time stationary property of speech signals,some successive frames are strong coherent.In this case,the current frame can use the previous' LP basis.Therefore adaptive LP basis is proposed.The experimental results show that the reconstruction performance by our proposed adaptive method is similar to that by non-adaptive way,but it is more economic on bandwidth and has better performance than that using DCT basis.