Audio Enhancement and Echo Cancellation

The Audio Enhancement and Echo Cancellation (AE + EC = AEP, Audio Enhancement Package)  is a high-performance, high-complexity system especially designed for high-end applications with demanding audio quality requirements. The algorithm consists of a series of modules which optimally compensate the undesired effects of echoes caused by acoustic feedback and reduce ambient noise, both inherent to hands-free devices. Typical applications are high-quality hands-free voice or video conferencing, high-quality car communication and remote surgery.

The particular principle of the algorithm is to model the echo path as a linear time varying system of Markow-Chain type and estimating it and the noise jointly with a Kalman Filter in the frequency domain. The advantages are superior performance than conventional EC and relatively low effort.

The AEP Algorithm has been heuristically improved, thus obtaining faster convergence in cases of moderate or fast echo path changes as is the case of automobile hands-free systems and called the new algorithm “Echo and Noise Enhancement Algorithm” (ENE). The post-filter subsystem in the diagram includes the Kalman Filter approach as part of a state-of-the-art single channel noise reduction algorithm.

The Equalizers and DC Filters are implemented as Wave Digital Filters thus having outstanding numerical properties and consequently SNRs, which allow the implementation with cost-effective fixed-point DSPs.

The center frequencies and attenuations of equalizer bands and the cut-off frequencies of DC filters are adjusted parametrically, which eases the setup of the AEP considerably.

The AEP code has been optimized for Blackfin but is available for a variety of DSP targets.