Efficient Framework for Compressed Sensing of Speech Signals

Compressed sensing has revolutionized several areas in digital signal processing, offering significant benefits in signal acquisition, processing, and reconstruction. Various frameworks have been proposed for the application of compressed sensing in speech processing. However, most of them lack efficiency and accuracy. In this paper, we present an efficient and accurate framework for the application of compressed sensing to speech signals. Our framework applies compressed sensing to thresholded discrete cosine transform (DCT) coefficients of overlapped speech frames. Then, we use adaptive reduced-set matching pursuit (ARMP) for speech reconstruction. Furthermore, we apply our proposed framework to speech enhancement. Our proposed framework achieves high reconstruction accuracy with very low complexity compared to other existing frameworks in literature.