Knowles Corporation, audio device solutions provider, announced the availability of the Knowles AISonic IA8201 Raspberry Pii Development Kit to bring voice, audio edge processing, and machine learning (ML) listening capabilities to devices and systems in a range of new industries.
“By selecting Raspberry Pi as the system host, we are opening up the ability to add voice and ML to the largest community of system developers that prefer a Linux or Android environment,” said Vikram Shrivastava, senior director, IoT Marketing at Knowles.
The new kit is built around the Knowles AISonic IA8201 Audio Edge Processor OpenDSP, for ultra-low power and high-performance for a plethora of audio processing needs. The audio edge processor combines two Tensilica-based, audio-centric DSP cores; one for high-power compute and AI/ML applications, and the other for very low-power, always-on processing of sensor inputs.
The IA8201 has 1MB of RAM on-chip that allows for high bandwidth processing of advanced, always-on contextually aware ML use-cases and memory to support multiple algorithms for an optimal user experience.
Using the Knowles open DSP platform, the kit includes a library of on-board audio algorithms and AI/ML libraries. Farfield audio applications can be built using the available ultra-low power voice wake, beamforming, custom keywords, and background noise elimination algorithms from Knowles algorithm partners such as Amazon Alexa, Sensory, Retune, and Alango to open up the design possibilities and ensure the freedom needed to support a wide range of voice and audio customisation.
The kit also features TensorFlow Lite Micro SDK for fast prototyping and product development for AI/ML applications. The TensorFlow Lite SDK allows for porting models developed in larger cloud Tensor Flow frameworks to an embedded platform at the edge, usually with limited compute and lower power consumption, for example, AI inference engines for anomaly detection in verticals such as industrial and commercial.
With options for either two or three pre-integrated Knowles Everest microphones based on product design needs, the kit includes two microphone array boards to help select the appropriate algorithm configurations for the end application, the company said. Developer support is available through the Knowles Solutions Portal for configuration
