Recently, multi-pitch detection which offers various information (such as number of simultaneous sounds, spectral envelopes, etc.) at the same time accompanied with s has been put stress for multi purposes, and numerous methods have been reported mainly in musical signal processing [1,2,3], speech signal processing [4,5] and auditory scene analysis [6,7].
Goto presented a method of tracking of objective single sound from polyphonic musical signals without restriction of the number of simultaneous sounds [1]. This method offers an optimal spectral envelope of the single sound by introducing a priori distrubution. Chazan et al. addressed a speech separation method by introducing a time warped signal model which allows a continuous pitch variations within a long analysis frame [5]. This method is prominent in respect of a capability of extracting not only the amplitudes of the partials but even the phases due to a parameter estimation of time domain signal model. Wu et al. described a multi-pitch tracking method in noisy environment by filter bank process and pitch tracking using HMM [4]. Although these methods actualize an accurate detection of s, either of them does not include specific process of detecting the number of speakers nor sounds as well as the majority of the previous methods.
Objective of our work is to detect s, respective spectral envelopes and a number of simulataneous speakers with a formulation of an optimal problem. A basic approach is stated in Section 2, and the detection algorithm as a whole is described in Section 3. And the results of operation experiments are reported in Section 4.