Ceva’s DSP and speech recognition integrate with TensorFlow Lite for Microcontrollers
Wireless connectivity and smart sensing technology provider Ceva’s DSP (digital signal processor) and speech recognition software for conversational AI and contextual awareness have been integrated with Google’s TensorFlow Lite for Microcontrollers, the company announced.
TensorFlow Lite for Microcontrollers is a cross-platform framework to implement tiny machine learning algorithms on power-efficient processors at the edge. Already available for Ceva-BX DSP cores, it expedites the adoption of low-power AI in conversational and contextual awareness applications.
“The increasing demand for on-device AI to augment contextual awareness and conversational AI workloads poses new challenges to the cost, performance and power efficiency of intelligent devices,” said Erez Bar-Niv, Chief Technology Officer at Ceva, in a prepared statement. “TensorFlow Lite for Microcontrollers dramatically simplifies the development of these devices, by providing a lean framework to deploy machine learning models on resource-constrained processors. With full optimization of this framework for our Ceva-BX DSPs and our WhisPro speech recognition models, we are lowering the entry barrier for SoC companies and OEMs to add intelligent sensing to their devices.”
Battery-operated IoT devices leverage tiny machine learning to deploy AI at the edge for on-device sensor data analytics in areas such as audio, voice, image and motion. Customers can thus benefit from a unified processor architecture to run both the framework and the associated neural network workloads.
The WhisPro speech recognition software is embedded in the TensorFlow Lite framework for voice-enabled IoT devices.
“Ceva has been at the forefront of machine learning and neural networks inferencing for embedded systems and understands that the future of ML is Tiny going into extremely power and cost constrained devices,” said Pete Warden, Technical Lead of TensorFlow at Google, in a prepared statement. “Their continued investment into powerful architectures, tools and software which support TensorFlow models provide a compelling offering for a new generation of intelligent embedded devices to harness the power of AI.”
Ceva-BX DSP is a programmable hybrid DSP and controller for various signal processing and control workloads of real-time applications. With an 11-stage pipeline and 5-way VLIW micro-architecture, it delivers parallel processing with dual scalar compute engines, load, store and program control that reaches a CoreMark per MHz score of 5.5, the company says. It provides support for SIMD instructions which makes it compatible with different signal processing applications and able to handle contextual awareness and sensor fusion algorithms with a wide dynamic range. Front-end voice, sensor fusion, audio processing, and general DSP workloads can be processed at the same time with AI runtime inferencing.
Last year, Ceva launched its second-generation AI processor architecture, NeuPro-S, to support deep neural network inferencing at the network edge.