Superior Voice Quality
The BT100 delivers reliable AVS performance and superior voice quality with the use of high-quality wideband audio codecs. It also acts as a superior hands-free speakerphone through advanced acoustic processing, and tailored audio streams optimized for speech recognition and two-way voice communication.
High-Quality Wideband Codec
The BT100 supports the high-quality narrowband and wideband Opus codec which gives significantly superior voice quality at comparable bit rates to other codecs.
From the left graph above, it is clear that the Opus codec in narrowband mode operating at 11 kbits/sec is equivalent to uncoded 3.5kHz low-pass filtered speech (ie no coding involved) while, from the right graph, the Opus codec in wideband mode at 20 kbits/sec is equivalent to 7kHz low-pass filtered speech. In addition to the Opus codec, the BT100 also supports G.711 (mu-law and A-law), G.729, G.722 (wideband ADPCM) as well as iLBC.
To maximize performance in degraded networks, the BT100 also employs network impairment compensation techniques such as a dynamic jitter buffer, packet loss concealment, and NetEQTM to mitigate impairments such as network jitter and packet loss.
Advanced Acoustic Processing
The BT100 is equipped with three digital microphones and uses advanced acoustic processing schemes such as linear & non-linear acoustic echo cancellation, non-linear processing, adaptive noise reduction, and beamforming to offer business-grade voice quality for the hands-free mode:
- Linear AEC: Adaptive linear acoustic echo canceler. Linear processing is suitable for local wakeword detection on the device and Amazon AVS speech recognition, and allows for the largest of room sizes with 256 ms of echo cancellation tail at 16kHz sampling.
- Nonlinear AEC (NLAEC): Suppresses nonlinearities due to, for example, saturation or mechanical nonlinear coupling. Designed for 2-way human communication to minimize the level of any audio echo.
- Nonlinear Processing (NLP): Uses non-linear processing techniques to replace any residual echo.
- Double talk detection: Detects presence of simultaneous near-end and far-end speech. Once Double Talk is detected, adjust the AEC to minimize convergence issues.
- Narrow-band Signal detection: Disables AEC adaptation if narrow-band signals are detected, as these signals can affect AEC convergence.
- Far-field Performance: Combines the three mic inputs to isolate speech from other background noise audio sources and so maximize performance even at distances of 15-20 ft from the device.
- Beamformer: Continuously tracks the direction of the Local Talker’s speech and hones in on their voice, increasing gain to the user while suppressing sound from other directions.
- Noise Reduction: Active noise reduction reduces the level of extraneous background noise without affecting speech quality.
- De-Reverberation (DRV): Removes the effect of reverberation introduced by audio reflections in the BT100’s local environment.
- Compressor / Limiter / Expander: Integrated system for boosting faint speech signals and attenuating and limiting loud signals to improve speech intelligibility from sources at a range of distances from the device.