Expanding on its leading edge AI solutions, Syntiant has recently announced the availability of a new reference design created in collaboration with Shanghai-based Telink Semiconductor. The production-ready reference design includes the Syntiant NDP120 Neural Decision Processor and Telink’s TLSR9 Series SoC, which features dual-mode compatibility with Bluetooth LE and classic Bluetooth.
"Our collaboration with Syntiant brings to market a full, ready-to-use, multilingual reference design, enabling manufacturers to quickly and efficiently deploy hands free voice interface capabilities in smart remote controls," says Wenjun Sheng, CEO at Telink Semiconductor. "Syntiant’s low-power deep learning architecture along with our Bluetooth solutions deliver greater functionality and performance at an attractive price point, while optimizing battery life."
With multilingual support including Mandarin and English, the remote control reference design is now available and suitable for manufacturers of media streaming devices, including smart TVs and set-top boxes, home entertainment and gaming systems, as well as smart home applications.
"We continue to see an increase in demand for voice functionality as people rather use hands-free speech to control their devices," says Kurt Busch, CEO of Syntiant. "However, sending wake words and commands to the cloud for processing is expensive and inefficient. With our neural decision processors and edge-optimized deep learning models, we can perform the AI processing required on the device itself, eliminating the cloud congestion and associated costs. Our collaboration with Telink has enabled us to provide a compact, highly accurate voice interface that consumes minimal power."
Syntiant’s suite of edge-optimized hardware and software solutions enables ODMs to quickly and easily add intelligence to almost any device for a variety of always-on voice, sensor and image applications, providing manufacturers with a natural interface free from cloud connectivity, ensuring privacy and security. The company’s data platform is designed to ingest data from crowdsourced or proprietary sources, expand the data set with synthetic data generation and produce high quality datasets for training.
Murata Type2DA Neural Decision Processor Module
Murata will be showcasing its new Edge AI module Type2DA with Syntiant NDP102 Neural Decision Processor (NDP) at CES 2023 (LVCC booth #10249). Ideal for always-on audio and sensor applications, the new solution significantly improves efficiencies while simplifying design challenges for a seamless integration and faster time-to-market.
This new Edge AI module provides a self-contained solution for audio and sensor processing and features an onboard hardware neural network. It works seamlessly with popular machine learning frameworks such as TensorFlow and the innovative codeless AI design suite from Edge Impulse. The on-board M4 host MCU and 512KB flash memory handle I/O responsibilities and streamline neural network (NN) model updates to enhance performance during and after deployment. With ultra-low current consumption (under 200 μW @ Vcore = 0.85v), the module can operate as a standalone decision unit or as the always-on component of a larger system.
The solution is available as a 45-pad LGA with a package size of 6.1 x 5.8 x 1.45mm. It is suitable for a variety of edge AI applications such as wake word and audio event detection, pressure sensing, gesture recognition, sensor fusion and other condition-based monitoring uses. For audio applications, the module offers a two-second sample buffer with hardware audio feature extraction, Mel scale conversion, and up to 100Hz frame rate. The embedded neural processor architecture provides direct neural network access to a 5-layer fully connected network. A Training Development Kit (TDK) is also available for developers training on neural network applications.
The Type2DA interfaces smoothly with Murata’s family of wireless modules and sensors, providing manufacturers a complete solution for low-power connected applications such as smart home and smart appliances. The Syntiant NDP102 delivers 100x improvement in efficiency and performance compared to current MCU-based offerings, and is capable of performing highly accurate sensor processing in a range of edge devices at under 100 microwatts.
"Our NDP102 provides an easily deployed machine learning solution for always-on audio or sensor processing in even the smallest of devices," says Mallik Moturi, chief business officer at Syntiant. "We designed the NDP102 with Murata in mind to bring powerful inference capabilities to all kinds of sensor applications across consumer and industrial use cases with near-zero power consumption."
At CES 2023, Murata will demonstrate sample applications with the Type2DA EVK, incorporating Wi-Fi connectivity and microphone and accelerometer sensors for easy evaluation, training data collection, and POC development. Syntiant will demonstrate its end-to-end deep learning solutions for always-on vision, audio and sensing applications at the event.
www.telink-semi.com
www.murata.com
www.syntiant.com
