XMOS has released its xcore-voice solution, claiming cost-effective voice interfaces without an audio quality penalty. The solution, built on the xcore.ai AIoT platform, XMOS’s third-generation xcore architecture, delivers voice pipeline example designs using the XMOS audio front-end.
(Source: XMOS)
The designs incorporate far-field processing and support for third-party ISV voice algorithms. The example designs leverage the processor capabilities through the xcore platform software, which provides the building blocks for a range of applications, said XMOS.
The software-defined, low-power voice solution addresses multiple markets using high performance DSP algorithms to extract human speech from noisy environments at a distance and running AI payloads including wake words and speech recognition. Xcore-voice can be used in home electronics applications such as appliances, smart TVs and soundbars, smart lighting and power switches, fitness machines, consumer toys and smart home gateways.
The xcore platform acts as a voice processor, freeing up the voice processing workload from the SoC, so that designers can achieve optimal audio quality, delivering voice interfaces fast and cost effectively, according to the company.
The platform focuses on three key areas: power, performance and flexibility. The xcore-voice combines always-on power of <55 mW, which is in line with the EU’s energy-saving regulations, high performance via a far-field voice algorithm in a package that enables customization and feature upgrades, application-specific components such as tunable algorithms and core application software components such as FreeRTOS.
It also delivers the xcore architecture flexibility, which offers general-purpose DSP, AI and IO processing that can be designed in software, reducing time-to-market for custom silicon. Although integration pathways are simplified, the solution is BOM-optimized for turnkey integration, said XMOS.
Features also include two PDM microphone interfaces, a digital signal processing pipeline; full duplex, stereo acoustic echo cancellation (AEC); point noise suppression via an interference canceller; flexible audio-output routing and filtering; a switchable stationary noise suppressor and programmable automatic gain control.
The xcore-voice platform delivers a natural voice interface for local command recognition, minimizing the need for the cloud, said XMOS. The result is lower latency, enhanced privacy, further cost savings and reduced power requirements.
The xcore-voice solution is currently available. The voice evaluation kit and explorer board evaluation kit, XK-VOICE-L71 and XK-EVK-XU316, respectively, are available via Digi-Key.
Advertisement
Learn more about XMOS
