Detection of phrase boundaries in Japanese by low-pass filtering of fundamental frequency contours

Major syntactic boundaries are often accompanied by a rise in the phrase component of the fundamental frequency (F₀) contour. Detecting such rises, therefore, can be significantly helpful to the speech recognition process. We developed a method to detect syntactic boundaries with phrase component rise (henceforth, phrase boundaries), based on the compression of the accent component of the F₀ contour (in logarithmic scale), using a low pass filter. In this method, F₀ contours are viewed as signals in the time domain, which can be roughly separated into phrase and accent components due to their different frequency contents. Phrase boundaries are detected whenever a significant rise occurs in the derivative of the filtered F₀ contour (H. Fujisaki a K. Hirose, 1984). The method managed to detect about 77% of manually detectable phrase boundaries, though with a relatively high insertion rate. The insertion rate can be reduced by using the partial AbS method, proposed by the authors (K. Hirose and A. Sakurai, 1996)