Multicore audio processor chip has machine-learning aptitude
The IA8508 is optimized for machine learning through DNN (deep neural network) hardware acceleration and an instruction set that is based on a 16-way SIMD (single instruction, multiple data) architecture. According to the manufacturer, the processor can perform specific machine-learning tasks 10 times more efficiently than competitive devices. Lightweight floating-point operations maintain audio signal fidelity, while minimizing power consumption.
To address issues related to noise and distance, the IA8508 uses algorithms for acoustic echo cancellation, dynamic beam forming, steering, sound classification, and noise suppression. Its low-power core is able to listen to a variety of key phrases concurrently. The part provides eight PDM (pulse density modulation) microphone interfaces, six bidirectional multichannel I2S/TDM links, and a PDM output. Processor interfaces include SPI, I2C, UART, and GPIOs.
The IA8508 is based on an open DSP platform, bringing together many contributors to the intelligent voice ecosystem. Knowles offers a full developer platform, including a software development kit, a plugin-based framework, example plugins, an evaluation module, compiler toolchain, test applications, optimized DSP toolkit libraries, and documentation.
Pricing for the IA8508 audio processor varies based on volume.
IA8508 product page
Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.