Instantaneous frequency and detection of dynamics in speech

摘要

We show that instantaneous frequency (IF) can be used for automatic detection of speech dynamics. A very simple procedure for IF estimation is developed which spends much less computer resources in comparison with standard IF calculation methods. The input file is divided into frames, and IFs are calculated at points related to local maximums in frame. All inferences about presence of speech dynamics in the file are done on the base of distributions of IFs inside of the frames. Efficiency of considered technique is demonstrated by the problem of source separation.