Improving speech intelligibility in in-vehicle audio systems under severe noise conditions using bark domain formant shifting

Abstract

During vehicle operation, the interior noise environment typically exhibits nonstationary and time-varying characteristics. Elevated background noise levels can significantly impair the intelligibility of audio systems such as navigation and voice broadcasts, thereby hindering the driver’s understanding of voice information and increasing potential driving risks. This paper proposes a speech signal preprocessing method that incorporates a novel Bark-domain formant-shifting technique and speech enhancement to improve intelligibility under complex and dynamic driving conditions. To address potential high-noise environments, an adaptive gain control algorithm is incorporated, thereby enhancing the effectiveness of intelligibility improvement under such conditions. Vehicle tests were conducted using a digital artificial head under various driving conditions and road environments. The performance and stability of the proposed method were verified using objective evaluation metrics. Additionally, a subjective listening test was conducted to complement the objective analysis. Both objective and subjective evaluation results indicate that the proposed method significantly improves the intelligibility of in-vehicle audio systems in severe noise conditions.