Echo canceller with reduced requirement for processing power

Abstract

An echo canceller processing echo, noise and near end talk in a narrower, but still intelligible, frequency band for reducing required processing power and complexity. In a preferred embodiment of the present invention, an input audio signal of captured sound in an audio communication system is decimated and then divided into a number of sub bands by an analyze filter. Each sub band is processed as in background audio echo cancelling by subtracting the signal with an echo estimate from a model of the acoustic signal in the respective sub band, except from that the signal is also bypassed, adjusted by a filter and subtracted from the processed signal. The resulting signals are then recombined by a synthesize filter and interpolated to the original sampling rate and bandwidth. Finally, the output from the synthesize filter is added to the input audio signal, which has been delayed and adjusted by a filter. The filters are controlled by a control algorithm detecting the presence of near end sound, far end sound and noise, so that the filters, and consequently the high pass filter of the echo canceller, only pass high frequency (above low pass frequencies) when only near end sound is detected.

Description

field of invention [0001] The present invention relates to audio communication systems and methods having improved acoustic properties, and more particularly to video conferencing systems including improved audio echo cancellation systems. Background of the invention [0002] In traditional conferencing system configurations using loudspeakers, two or more communication components are placed at different locations. Transmitting a signal from one location to another using a conferencing system experiences several delays, including transmission delays and processing delays. For a video conferencing system, the processing delay of the video signal is quite large compared with the processing delay of the audio signal. Since the video and audio signals are presented synchronously and in phase, a lip-sync delay is purposefully added to the audio signal in the transmit and receive signal paths in order to compensate for the longer video signal delay. [0003] In a conventional co...

AI Technical Summary

Problems solved by technology

It is possible to reduce the filtering / adaptation part by increasing the number of subbands, but at the cost of increasing the overhead of computing the subband signals

However, by using many subbands, i.e. using a large number of fast transforms, it is possible to obtain a fullband * log 2 Fs fullband added complexity

[0021] Although the above approach is theoretically possible, it can be difficult to implement in practice due to cache invalidation during signal processing when implementing a large number of transformations

Images