Signal Representation Including Waveform Envelope by Clustered Line-Spectrum Modeling
Michiko Kazama, Kazuaki Yoshida, Mikio Tohyama · Journal of the Audio Engineering Society · 2003
Clustered line-spectrum modeling (CLSM) around the peaks of interpolated FFT spectrum records has been developed for signal analysis and representation including the signal envelope. The signal waveform, including the envelope, can be represented by components clustered around spectrum peaks. The sinusoidal components at the peaks are estimated by peak picking, whereas the clustered components required in particular for envelope representation can be estimated by obtaining least-squares error (LSE) solutions in the frequency domain, which was originally formulated by Quatieri and Danisewicz and by Maher. Numerical simulation reveals that the basic CLSM algorithm works well, and a narrow-band speech sample or impulse-response type transient-signal analysis shows that acoustic signals that include envelopes can be expressed quite effectively by using clustered components based on CLSM.