The multi-lag-window method for robust extended-range F/sub 0/ determination
Edouard Geoffrois · 2002
The paper addresses the problem of the fundamental frequency (F/sub 0/) determination of a speech signal, and proposes four improvements to conventional frequency-domain methods. The major improvement is a multi-scale analysis which extends the range of F/sub 0/ that can be correctly processed. It builds on the lag-window method proposed by Sagayama (1978), hence the name "multi-lag-window". Secondly, a modification of the lag-window method itself improves its robustness to periodic noises (while loosing its gain-independence property). Thirdly, a rescaling is introduced to permit a full dynamic programming search for the optimal F/sub 0/ curve. Finally, a mathematically justified peak interpolation is proposed for replacing the conventional, inaccurate parabolic interpolation. These four improvements result in an accurate, robust, extended-range F/sub 0/ determination method, which was tested on spontaneous speech from 20 speakers, ranging from less than 50 Hz to more than 600 Hz.