CouthIT’s acoustic echo cancellation implementation is designed to continuously and adaptively cancels the unwanted echo generated by the coupling between the loudspeaker and the microphone and maximize the user experience by providing high quality full duplex communication in all environments. The algorithm implemented is compatible with the ITU-T G.167 recommendation. The implementation is targeted for wide range of applications such as VoIP, tele-conferencing and loud speaking audio systems.
- Fixed-point ANSI C implementation.
- Re-entrant implementation.
- C-callable APIs
- Module is fully interruptible.
- Implementation independent of the sampling rate.
- Operates on a 10ms frame size.
- Fast Convergence rate (greater than 20dB/sec).
- Automatically adapts to the real time echo path changes
- User can configure to use either full duplex or near full duplex mode depending on hardware capability and requirement.
- Robust novel double talk detector based on speech detectors and cross correlation statistics to prevent loss of convergence.
- Fully compliant to ITU-T G.167 requirements.
- User configurable echo tail length in a multiple of 10ms upto 120ms.
- Protection against non linearity caused by low frequency components.
- Provides more than 60dB of acoustic echo cancellation.
- Howling control to eliminates acoustic howling / squealing interference.
- Effective non linear processor to remove non-linear residual echo with smoothing.
- High pass filter to remove non-linear low frequency components thereby improving the performance of AEC.
- Insertion of comfort noise during silence periods to discontinuity in the speech signals.
- Optimized for low memory foot print and low complexity.
- Integrated with AGC to maintain dynamic range of the speech signal (optional).
- Integrated with NS to remove residual echo and background noise (optional).
- Tested using a large database of male and female speech test vectors.
- Tested for full duplex operation and for divergence in double talk scenario.
- Tested for removal of howling / squealing interference.
- Tested for convergence in case of high level of background noise.
- Tested for convergence in the presence of varying speech signal levels.
- Tested in conjunction with CouthIT’s Noise suppression, AGC, jitter buffer and speech codec modules.
- Tested for graceful exit in case of errors or exception.
- Tested for Input buffer corruption
- Tested for I/O buffer alignment requirements
- Tested for multi-instance implementation.
- Tested for 100% code coverage
- Range validation of all API parameters.
- Code validated on ARM926EJ-S (OMAP L138) platform running Open Embedded Linux
ARM9E, ARM11, Cortex-A8, and Cortex-A9