The Implementation and Optimization of AMR on Mobile Device

Abstract

Considering the limited resource of mobile device, the operation complexity of the applications needs to be reduced. This paper presents the implementation of Adaptive Multi Rate (AMR) on mobile device and especially focuses on reducing the computational complexity. AMR is an algebraic code excited linear prediction (ACELP) based algorithm which could achieve satisfying speech quality by using the analysis-by-synthesis method. It has been chosen by the Third Generation Partnership Project (3GPP) as the mandatory codec for the third generation (3G) cellular systems. The optimizations include algorithm and instruction level. Based on the mobile device i-mate JAMin which has the ARM processor operating at 195 MHz, one second speech costs only 0.6 second processing time. So, more resource is reserved for other applications.