In mobile communications, speech is often degraded by ambient noise and acoustic echo. Both problems have attracted a high level of research interest over the last decades. Recently dual channel solutions for noise reduction have been investigated and can yield better
performance than single channel solutions. This paper presents an analysis of the echo problem based on recordings with a handsfree dual channel mock-up phone and proposes a novel residual echo power spectral density (PSD) estimator that uses two microphone
signals instead of one. The proposed PSD estimate is compared to an existing single channel residual echo PSD estimator before being assessed within an echo postfilter. Our experiments show that the proposed dual-channel echo PSD outperforms the single channel
postfilter, especially at low signal to echo ratios.
