XMOS launches voice processors for smart devices

Update: August 6, 2023

They are the XVF3610 (demo board right), and an Amazon wake-word-enabled variant called XVF3615.

According to XMOS, Amazon’s Alexa includes the earlier XVF3510.

“The designs mark an evolution of the XVF3510 voice, and provide audio echo cancellation and interference cancellation algorithms to suppress noise from dominant sources and enable cross-room barge-in,” it said. “XVF3610 and XVF3615 also include automatic audio reference delay calibration – a critical capability for voice enabled TV systems.” – barge-in is where a person speaks over amplified audio to get the attention of the processor.

The Amazon-enabled version is for single pass Alexa-enabled systems, allowing manufacturers to off-load wake-word processing from their host processor.

XVF3610 has two firmware variants:

  • XVF3610-UA for USB accessory devices – audio and control via a USB2.0 interface
  • XVF3610-INT to be built into products – audio via I2S and control over I2C interfaces (right)

At the same time as the two ICs, XMOS announced a customisable design called Avona for voice, wake-word, speech-to-intent and intelligent voice applications that “allows engineers to incorporate keyword, event detection or local dictionary support to create a voice interface solution”, according to XMOS.

Alongside this is the xcore software development kit which incorporates a FreeRTOS environment for user application code and can be programmed in C or C++. It also supports bare-metal use.

Avona is scheduled to be available early in 2022.

The Amazon-enabled XVF3615  is scheduled to be available in January 2022.

The XVF3610 Voice Processor data sheet is here