A Correcting Method for Frequency Changes in Speech Speed Converting Process.
Masashi Kawasumi, Takafumi Ohnishi, T. Shimura, Keiji Hiramatsu · Journal of Life Support Engineering · 1999
In order to listen to recorded speeches quickly or slowly(compared to normal speed), it is necessary to change the playback speed of player. Since the speech wavefbrms are compressed or expanded in time axis according to the playback speed, this operation obviously causes damage to tone quality of voice, individuality of speaker, and understanding of the recorded speech. To fix the quality of the affbcted speech, there are two approaches, namely, cutting or pasting of these damaged wavefbrms in time domain and correcting the shi fted tone in frequency domain. The latter approach has advantages in holding frequency features of the speech through the process. Weh have been investigating a method of the approachin which the instantaneous frequency of signal is modifiable to any value. In this paper, this method is introduced and simulation results are discussed. In the results, both individuality of speaker and understanding of the recorded speech are almost fixed.