Smart microphone array design for speech enhancement in financial VR and AR
Rui Ma, Guocheng Liu, Qi Hao, Cong Wang · 2017
Signal quality is critical in speech recognition during Virtual Reality/Augmented Reality (VR/AR) based interactions, especially in financial applications. This paper presents a 6-element concave microphone array designed for the VR/AR headset. Two key processing methods are developed: (1) beamforming technique to enhance the signal along the target speech direction, and (2) ICA blind source separation to separate mixed speech signals. These two methods are further integrated to improve each other's performance and to achieve an intelligent target speech enhancement scheme. Different speech interactions in VR/AR financial transactions are investigated with the smart microphone array. Simulation and experiment results have demonstrated that the proposed system can improve SNR and yield a smaller WER (word error rate) in speech recognition.