Title :
On the duality between line-spectral frequencies and zero-crossings of signals
Author :
Kumaresan, Ramdas ; Wang, Yadong
Author_Institution :
Dept. of Electr. Eng., Rhode Island Univ., Kingston, RI, USA
fDate :
5/1/2001 12:00:00 AM
Abstract :
Line spectrum pairs (LSPs) are the roots (located in the complex-frequency or z-plane) of symmetric and antisymmetric polynomials synthesized using a linear prediction (LPC) polynomial. The angles of these roots, known as line-spectral frequencies (LSFs), implicitly represent the LPC polynomial and hence the spectral envelope of the underlying signal. By exploiting the duality between the time and frequency domains, we define analogous polynomials in the complex-time variable ζ. The angles of the roots of these polynomials in ζ-plane now correspond to zero-crossing time instants. Analogous to the fact that the line-spectral frequencies represent the spectral envelope of a signal, these zero-crossing locations can be used to represent the temporal envelope of bandpass signals
Keywords :
polynomials; signal representation; spectral analysis; speech processing; LPC polynomial; angles; antisymmetric polynomials; bandpass filtered speech signal; complex-frequency plane; complex-time variable; frequency domain; line spectrum pairs; line-spectral frequencies; linear prediction polynomial; signal zero-crossings; spectral envelope; symmetric polynomials; temporal envelope representation; time domain; z-plane; zero-crossing time instants; Band pass filters; Frequency domain analysis; Frequency synthesizers; Linear predictive coding; Polynomials; Quantization; Signal synthesis; Speech coding; Speech recognition; Speech synthesis;
Journal_Title :
Speech and Audio Processing, IEEE Transactions on