Sound Source Localization and Speech Enhancement Algorithm Based on Fixed Beamforming

Abstract

In voice-based human-computer interaction, various environmental noise may interfere with the normal transmission of information. Noises may typically lead to performance degradation and efficiency reduction of the voice interaction system. Aiming for reliable sound source localization and effective speech enhancement, we researched those algorithms in scenario of voice-based human-computer interaction with indoor intelligent robots, carried out analysis and implementation. In this paper, we analyzed the localization algorithms based on steered-response power, improved, implemented, and tested on it for its stability in practical application. The results indicated the good performance of the algorithm we proposed under the condition that no strong directional noise interferes with the target signal. Reliable two-dimensional sound localization information can be provided, ensuring real-time performance of the system. We also implemented the speech enhancement algorithm based on fixed beamforming, using the sound localization information provided by the two-dimensional sound localization algorithm. The results showed that the algorithm has a certain inhibitory effect on non-correlative noise on the premise of reliable sound source localization and undistorted signal of speech.