SPEECH REINFORCEMENT BASED ON BINAURAL LOUDNESS MODEL

This paper proposes a speech reinforcement system based on a binaural loudness model to improve intelligibility of the speech in noisy communication environments. Compared to conventional monaural-based approaches, the proposed method provides more suitable gains to a speech signal because the binaural model effectively reflects the influence of noise on both ears. The optimization algorithm based on the golden section search and parabolic interpolation is applied to solve a minimization problem needed for gain calculation. Test results confirmed that the proposed system notably enhanced the clearness of target speech while maintaining speech quality and providing similar loudness with the clean speech.