基于启发式掩模 EMD 的音频突变成分检测方法

Under many conditions, audio signals could be consisted of composed of steady components and mutant components these two components and there was a clear difference between steady components and mutant components on features. The mutant components usually carried more important information, which was the target of signal processing to analyze. To detect mutant components and separate these two components effectively, it was necessary to accurately detect and extract the mutant components. Thus, the paper proposed an audio mutant component detection method based on heuristic EMD with a masking signal. In this detection method, heuristic EMD with a masking signal was used to decompose the audio signal and extract the instantaneous information of each point as the detection feature, and meanwhile a window adaptive updating strategy was proposed in this paper to set the suitable lengths of mutant components. In IOWA’s instrument audio dataset, this detection method could detect audio mutant components with a detection accuracy of 98.68% and a detection recall of 87.65%. What’s more, the detection method does not require human intervention and the detected mutant components have the same dimensions, which is convenient to perform subsequent processes such as feature extraction, classification and so on.