xcore.ai’s unique architecture combined with the XCORE-VOICE framework allows designers to incorporate keyword, event detection, or advanced local dictionary support to create a complete voice interface solution.
XCORE-VOICE is a highly flexible voice solution that allows you to pick and choose voice components to run alongside your desired AI model. Detailed documentation is available to guide you through options as highlighted in the sections below.
We have included two example designs that will enable you to get started immediately.
Following that you can download the source code to further compile, according to your needs.
Far-Field Voice Local Command
- Dual PDM microphone with point and stationary noise reduction and keyword spotter.
- Recognising 20 sample commands Allows control of peripherals and audio play-back “spoken response” option. System integration via UART.
- Command dictionary easily customised though our partners.
Far-Field Voice Assistant
- 2-microphone far-field voice assistant front-end
- Audio pipeline including echo cancelation, interference cancelling and noise suppression
- Stereo reference input and voice assistant output each supported as I2S or USB (UAC2.0)
An XMOS solution is a convenient way to access a collection of software, examples, libraries and documentation for developers with an application in mind, and where all the components are guaranteed to work together.
XMOS Frameworks are a collection of high-level components, already interoperable and sharing a common design approach.
|A collection of audio and voice signal processing components, implemented to use the 32-bit Block-Floating Point arithmetic of the xcore.ai Vector Processing Unit (VPU)||HTML|
|CONTAINED WITHIN FWK-VOICE||DESCRIPTION||DOCUMENTATION|
|Acoustic Echo Canceller (AEC)||Configurable for number of microphones and play-back channels||HTML|
|Automatic Delay Estimation and Correction (ADEC)||Functions for measuring and correcting delay offsets between AEC reference and loudspeaker signals||HTML|
|Interference Canceller (IC)||Point noise cancellation, using two microphone input||HTML|
|Noise Suppression (NS)||Removes stationary noise||HTML|
|Automatic Gain Control (AGC)||Maintains output between target levels||HTML|
|Voice To Noise Ratio Estimator (VNR)||ML based alternative to a classic VAD, providing a more informative estimation of voice presence in noise.||HTML|
|XMOS’s SMP port of FreeRTOS||HTML|
XMOS Libraries provide reusable software, as source, for common functions
|Allows 1 to 16 PDM microphones to be connected to an xcore device, with configurable filtering and decimation.||HTML|
|Comprehensive library of mathematical and DSP functions, including integer, float, vectorised and complex operations.||HTML|