Amazon qualifies XMOS Alexa development kit

“Amazon is seen by many as the leader in the world of voice, and our engineering expertise will be helping to transform the way consumers interact with everyday devices like TVs,” claimed XMOS CEO Mark Lippett.

The kit is based on XMOS’ XVF3510 voice processor (mid PCB in photo) and includes a Micro-USB connector for power and USB device connectivity, a “low-jitter audio-quality clock”, according to XMOS, line-in port and xTAG interface for firmware updates.

The dual microphones come on a daughter board (right-most in photo) and are Infineon IM69D130 PDM (pulse density modulation) MEMS types with 71mm inter-mic spacing.

The kit needs Requires Raspberry Pi 3 to run AVS SDK and the Sensory TrulyHandsfree wake-word engine.

To connect to this, the kit includes a suitable Raspberry Pi HAT (on Raspberry Pi in photo), which also has I2S audio and I2C connectivity, a 24bit stereo DAC, four user-controlled buttons and two LEDs.

The XVF3510 voice processor runs far-field voice capture algorithms including: full duplex stereo acoustic echo cancellation (AEC), barge-in, the two channel digital microphone interface, interference cancellation, noise suppression, automatic delay estimator control (ADEC) and automatic gain control (AGC).

The interference canceller nulls point noise to cancel out typical noise in the soundscape, stereo acoustic echo cancellation suppresses unwanted speaker echo and enables barge-in across audio playing from the device, while noise suppression removes background noise from other devices.

“Our automatic delay estimator adjusts audio reference signal latency dynamically, ensuring the acoustic echo cancellation algorithms deliver a smooth, real-time barge-in experience,” according to XMOS.

An Alexa developer account is also required.