Some fast speech processing algorithms using AltiVec technology

Author

Joshi, Sanjay M. ; Dubey, Pradeep K.

Author_Institution

CSEE Dept., Maryland Univ., Baltimore, MD, USA

Volume

4

fYear

1999

fDate

15-19 Mar 1999

Firstpage

2135

Abstract

The AltiVec technology is a SIMD (single instruction multiple data) extension to PowerPC architecture. It is intended to provide architectural support for performance improvement of various image and signal processing applications, including speech processing, on a general-purpose processor implementation, such as, the PowerPC line of processors. We have implemented some of the common speech processing algorithms on AltiVec architecture. The algorithms discussed in this paper are autocorrelation computation, linear prediction coefficients computation via Levinson-Durbin method and Schur (1917) recursion and part of the GSM speech compression system. AltiVec obtained significant speedups on all these algorithms, compared to the scalar PowerPC implementation. We also found that additional speedup was achievable by porting to new, more SIMD-friendly algorithm

Keywords

correlation methods; data compression; digital signal processing chips; parallel algorithms; parallel architectures; prediction theory; speech coding; AltiVec architecture; AltiVec technology; GSM speech compression system; Levinson-Durbin method; PowerPC architecture; PowerPC processors; SIMD-friendly algorithm; Schur recursion; autocorrelation computation; fast speech processing algorithms; general-purpose processor; image processing; linear prediction coefficients; performance; signal processing; single instruction multiple data; speech coding; speedups; Analytical models; Assembly; GSM; Parallel processing; Performance analysis; Registers; Signal processing algorithms; Speech processing; Table lookup; Trademarks;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1999. Proceedings., 1999 IEEE International Conference on

Conference_Location

Phoenix, AZ

ISSN

1520-6149

Print_ISBN

0-7803-5041-3

Type

conf

DOI

10.1109/ICASSP.1999.758356

Filename

758356