Sound Source Localization Based On Compressive Sensing With Microphone Array

For the past few years, multimedia technology develops fast, applications like radio teleconference and distant sound field reconstruction show us vast potential for future development. Sound source localization based on microphone array plays an important role in these applications. Sound sources’ locations determine the structure of sound field and they are the most important parameters for sound field’s sparsity. However, as signal’s bandwidth increases, the sampling rate and data storing cost become larger. As a result, the sound source localization system occupies more power with a lower efficiency.To solve the high sampling rate problem, a sound source localization algorithm based on Compressive Sensing. Compressive sensing is a newly proposed theory for signal processing. It concerns the sparse signal sampling and reconstruction. In the proposed algorithm, a microphone is selected as reference and it samples the sound field by Nyquist sampling rate. Other microphones are compressive sensors and sample the sound field by compressive rate which is much lower than the Nyquist sampling rate. By processing these signals and the TDOA(Time Difference of Arrival) method, we can derive the sound source location estimation. A fractional-delay filter is inserted after the referencing microphone, the algorithm can also locate off-grid sound sources satisfactorily.Theoretical analysis and experiments show that, the proposed algorithm can give a satisfying estimation of on-grid and off-grid sound sources with the error lower than 3cm in the sound field of 20m*30m. As TDOA has an advantage of anti-reflection, the location error in a reflecting sound field is also an important performance for the proposed algorithm. Simulation shows that the algorithm can work well in a half-free sound field with reflection. When the reflecting ratio is 0.5, the location error is smaller than 15 cm.