Evaluation of EVS, AMR-WB and AMR-NB Codecs Under Different Background Noise Scenarios

We evaluate the Enhanced Audio Services (EVS) codec for subjective quality under different background noise scenarios using listening tests and Perceptual Objective Listening Quality Analysis (POLQA). We compare the performance of EVS against other codecs used in mobile phones; Adaptive Multi-Rate Wideband (AMR-WB) and Adaptive Multi-rate Narrowband (AMR-NB). Listening tests were conducted with 50 users (25M, 25F, age range 20–33 years) using the earpiece and speaker of the mobile phone. In addition to clear speech, background sounds specific to daily Turkish life, including street markets, cafes, Turkish music and television, were played from a speaker in the room while the participants were speaking; users were asked to listen to the speech on the phone and award a mark in the range 0–10. EVS (9.1 st. dev. 1.2) was found to have similar quality for clear speech without background noise to AMR-WB (9.1 st.dev. 1.1). There was similar performance in the presence of background sound (e.g. market high) for EVS (6.6 st. dev. 2.2) and AMR-WB (6.6 st. dev. 2.1). A. The paired t-test comparing each codec with EVS showed little statistical significant difference between EVS and AMR-WB for all conditions. AMR-NB is always perceived poorly.

I. Introduction

Speech coding technology plays a crucial role in modern communication systems, enabling high-quality voice communication over limited bandwidth channels. The Enhanced Voice Services (EVS) codec (3GPP TS26) [1] is a recent speech coding technology developed for the 3rd Generation Partnership Project (3GPP). It is designed to provide superior performance compared to existing codecs, including Adaptive Multi-Rate Wideband (AMR-WB) and Adaptive Multi-Rate Narrowband (AMR-NB).