Parallel implementations of beamforming design and filtering for microphone array applications

One of the main limitations of microphone array algorithms for audio applications has been their high computational cost in real acoustic environments when real-time signal processing is absolutely required. Regarding audio/speech signal processing, beamforming algorithms have been used for the recovery of acoustic signals from their observations when they are corrupted by noise, reverberation and other interfering signals. In order to reduce their high computational load, frequency-based filtering have been used to achieve a real time application. Our research focuses on the use of different multicore/manycore platforms in order to achieve a real time beamforming application in the time domain. Efficient algorithms has been proposed and tested in several devices and results have shown that GPU implementation of beamforming design and filtering outperforms multicore implementation in computational cost terms. The performance obtained suggests that GPU implementation paves the way for low-cost real-time audio beamforming applications.