Journal of Applied Sciences

• Articles • Previous Articles     Next Articles

Optimization Design for Efficient Computation of Line Spectrum Pairs

LI Juan-juan1, YU Yi-biao2, RUI Xian-yi3   

  1. 1 School of Mechanical and Electronic Engineering, Zhejiang Ocean University, Zhoushan 316004,China
    2 School of Electronic Information Engineering, Soochow University, Suzhou 215021,China
    3 Department of Electronic Engineering, Shanghai Jiaotong University, Shanghai 200030,China
  • Received:2006-06-24 Revised:2006-10-30 Online:2007-05-31 Published:2007-05-31

Abstract: This paper proposes a new algorithm APF for calculating line spectrum pairs (LSP) of speech signal. LSP parameters can be obtained by solving an N-order non-linear equation relating to the linear predictive coding (LPC) coefficients. First, one root of the N-order nonlinear equation is derived with the Aitken iterative method. The order is then decreased with polynomial synthesis division. Finally the fourth-order equation is solved using the Ferrari’s solution. Theoretical analysis and experiments on TMS320C5510 DSP show that the proposed algorithm has high precision and low computation complexity, meeting the real-time processing requirement. The algorithm can be applied to low-bit real-time speech coding such as G.729A systems.

Key words: line spectrum pairs, linear predictive coding, real-time processing