Sound source identification by compressive sensing

Abstract

Compressive sensing (CS) is an emergent technology that has found increasing application in the areas of broadband signal monitoring and image reconstruction [D. Donoho, IEEE Trans. Info. Theory 52, 1289–1306 (2006)]. It involves a novel approach to sampling in which the salient information in signals is recovered from the projection of observations onto a small set of randomly‐selected basis functions. The major attraction of CS is its capacity for accurate reconstruction based on sparse sampling and little or no prior knowledge of signals. This feature makes it attractive, as well, for application to the problem of sound source identification; an allied task in which sound waveforms as signals are to be classified according to their generating source. In the present paper CS is applied to examples of sound source identification tasks taken from the human psychoacoustics literature. The examples are used to demonstrate in these cases potential advantages of CS classification over traditional decision algorithms that sample at the Nyquist rate. Parallels to the human data are also noted, entertaining speculation as to the role CS classification might play in theoretical thinking about human sound source identification. [Research supported by NIDCD grant 5R01DC006875‐02.]