This study proposes a frame-based howling detection method.
The method uses a VAD algorithm in the pre-processing step.
Applying VAD reduces computational complexity and false alarm probability.
Multiple features of the howling are checked to increase the accuracy.