INTERSPEECH 2012
13th Annual Conference of the International Speech Communication Association

Portland, OR, USA
September 9-13, 2012

A Triple-Microphone Real-Time Speech Enhancement Algorithm Based on Approximate Array Analytical Solutions

Ryan Ritch, Meng Yu, Jack Xin

Department of Mathematics, University of California, Irvine, CA, USA

A novel triple microphone array speech enhancement algorithm is developed based on closed form solutions to an approximate algebraic system of equations on noise directivity and speech spectral energy. The system is derived from the ensemble averaged spectral energy equations of a first order differential microphone array. For sufficiently small microphone spacing, closed form analytical solutions are derived using suitable approximations for the normal speech frequency range. The resulting algorithm is simple to implement and efficient for real time noise reduction in a reverberant environment. The algorithm's limitation, the case of spatially almost overlapping sources, is analyzed mathematically, leading to an effective alternative solution utilizing array rotation. The algorithm is evaluated and compared with well-established beamforming algorithms. Results indicate robust gains in objective measures of speech relative to the baseline algorithms in the presence of multiple/mobile noise sources.

Index Terms: triple-microphone array, spectrum estimation, noise directivity, closed form solutions, real-time algorithm

Full Paper

Audio Examples
Explanation
Noisy    Enhanced   One noise (30)
Noisy    Enhanced   One noise (150)
Noisy    Enhanced   Two noises
Noisy    Enhanced   Three noises
Noisy    Enhanced   Four noises

Bibliographic reference.  Ritch, Ryan / Yu, Meng / Xin, Jack (2012): "A triple-microphone real-time speech enhancement algorithm based on approximate array analytical solutions", In INTERSPEECH-2012, 1942-1945.