Howling Detection and Gain Control for Speech Reinforcement in a Noisy Car Cabin Environment

In-car speech communication is particularly challenging due to environmental noise. The speaker's microphone also acquires car and road noises, resulting in a low signal-to-noise ratio and persistent frequency-howls that do not decrease, which degrade the system's output sound quality. In this paper, we address the problem of howling control for in-room speech reinforcement systems under high environmental noise levels, specifically in a car cabin. We control the acoustic feedback via a gain control algorithm based on howling detection. Two detection methods are proposed to detect non-decreasing underdamped frequency-howls. A single-channel optimal Feedback Wiener gain is derived to enhance the desired near-end speech signal, followed by another Wiener filter that is based on the signal magnitude relative to the estimated noise power spectral density. Adjusting the howling energy threshold is proposed to deal with false-detection artifacts arising as the environment's environmental noise level rises. The performance improvements of the howling-detection-based gain control algorithm following the proposed adjustments are evaluated in clean and noisy environments. As detection of non-decreasing underdamped frequency-howls is feasible, it was found that it becomes unnecessary when the environmental noise is successfully reduced via the Wiener filter.