Acoustic echo and noise canceller for personal hands-free video IP phone

Abstract

This paper presents implementation and evaluation of a proposed acoustic echo and noise canceller (AENC) for videotelephony-enabled personal hands-free Internet protocol (IP) phones. This canceller has the following features: noise-robust performance, low processing delay, and low computational complexity. The AENC employs an adaptive digital filter (ADF) and noise reduction (NR) methods that can effectively eliminate undesired acoustic echo and background noise included in a microphone signal even in a noisy environment. The ADF method uses the step-size control approach according to the level of disturbance such as background noise; it can minimize the effect of disturbance in a noisy environment. The NR method estimates the noise level under an assumption that the noise amplitude spectrum is constant in a short period, which cannot be applied to the amplitude spectrum of speech. In addition, this paper presents the method for decreasing the computational complexity of the ADF process without increasing the processing delay to make the processing suitable for real-time implementation. The experimental results demonstrate that the proposed AENC suppresses echo and noise sufficiently in a noisy environment; thus, resulting in natural-sounding speech.