This paper proposes a novel howling detection criterion for speech applications, which is based on temporal coherence (will be referred as TCHD). The proposed TCHD criterion is based on the fact that the speech only has a relatively short coherence time, while the coherence times of the true howling components are nearly infinite since the howling components are perfectly correlated with themselves for large delays. The proposed TCHD criterion is computationally efficient for two reasons. First, the fast Fourier transform (FFT) can be applied directly to compute the temporal coherence. Second, the proposed TCHD criterion does not need to identify spectral peaks from the raw periodogram of the microphone signal. Simulation and experimental results show the validity of the proposed TCHD criterion.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.