This blog post was originally published at Synaptics’ website. It is reprinted here with the permission of Synaptics.
The world of connected devices is an exciting and dynamic one as companies develop more sophisticated products to enhance our convenience, entertainment and safety. At the heart of this new generation of smart devices is an increasing ability to enrich them with AI. Thanks to ever-improving machine learning algorithms and neural networking technology, the way we interact with systems – and the information they can perceptively provide – results in a much more intuitive and immersive experience.
At Synaptics, we recognize the vast potential of AI-enriched IoT across all aspects of our lives – at home, work and on the go. We also know that “one size does not fit all” when it comes to the underlying silicon platforms that enable these intelligent devices. The wide range of applications and use cases that are possible demand a variety of options for performance, cost, power and size.
Our powerful and efficient AI-based neural network processors make Synaptics well-positioned to address these challenges and enable Edge-based functions that are in high demand for consumer IoT products. With our multimedia processors, features such as voice biometrics, video object and face recognition, gesture controls, predictive maintenance alerts and intuitive user preferences can be applied in smart devices that don’t depend on the bandwidth and security issues of connecting to the cloud.
An award-winning foundation
Our VS600 family is the foundation of our offering to drive innovation in consumer IoT. This highly integrated SoC platform combines the processing performance required for AI-enabled systems as well as the efficiency to operate on the Edge. Through innovative engineering, it integrates an NPU supporting trillions of operations per second, a quad-core CPU subsystem, a GPU, an ISP and video and display processing. It’s a true multi-modal solution, supporting voice, audio, vision and video, enabling just about any device to offer state of the art interface capabilities.
The first generation of this family, the VS680, was recently named the Embedded Processor of the Year in the prestigious Analyst Awards from The Linley Group. Synaptics joined select company in the 2020 winner’s circle, alongside the likes of Intel, Apple and nVidia. Senior analyst Mike Demler noted: “Synaptics’ VideoSmart VS680 is a multimedia powerhouse that packs multi-core video and audio processing with a power and energy-optimized neural network and a custom deep-learning accelerator to bring artificial intelligence to smart home devices. By making a cost-effective AI-based Edge solution, Synaptics is enabling a more human-like experience in the products that run our homes and lives.”
Awards are great, but we are especially proud of how our customers are leveraging this powerful platform in products as such as set-top boxes, smart speakers and displays, video sound bars, media streamers, and home and security systems.
The VS640 broadens options for IoT
Building on the same foundation, we recently introduced the VS640, giving product developers a lower cost option for consumer IoT devices. It’s particularly well-suited for an emerging segment of video products, offering support for the next generation AV1 video decode format — a requirement for future YouTube and Netflix content — and also is capable of supporting the latest version of all major operating system frameworks including Android TV, RDK, Linux and AOSP. This lets service providers more easily integrate applications and custom user interfaces of their choice.
Notably, the VS640 will support all the major Conditional Access Systems (CAS) including Nagra, Synamedia, Irdeto, Verimatrix and Digital Rights Management (DRM) protection including Widevine and PlayReady, while also supporting security for all the primary internet streaming services.
Open AI Development Tools
To provide higher levels of privacy, security and real-time interaction for end-users, our consumer IoT platform leverages SyNAP™ (Synaptics Neural Network Acceleration and Processing). The SyNAP toolkit is an open, full-stack solution that builds upon industry-standard AI frameworks such as TensorFlow, TensorFlow Lite, and ONNX to simplify the implementation of customized machine learning inferencing.
Because it shares a common architecture with the VS680, developers using the VS640 can leverage a common SDK to enable faster time-to-market and more efficient software development.
Partnerships add AI-enriched features
Synaptics’ open architecture and extensive development tools allow our SoCs to be optimized for a variety of applications. One such example is in the emerging area of voice biometrics, which can be implemented to enhance device security and personalization through speaker recognition.
We recently announced a partnership with ID R&D, considered to be the leader in voice biometrics software, which results in the first deployment of voice biometrics on an NPU for smart home applications. By allowing the ID R&D software to tap into the Synaptics NPU, the solution enables developers to offer highly accurate AI-based voice biometrics on an extremely small footprint – ideal for products like smart speakers, set-top boxes, displays and security systems.
Leveraging the capabilities of the Synaptics NPU allows the inferencing component of the voice biometric capabilities to be moved off the CPU where they traditionally reside and results in more than a 10x speed improvement, a reduction of CPU utilization by a factor of three, and a significant reduction in RAM requirements.
John Amein, SVP at ID R&D noted, “Previous systems have been limited by the capacity of the core CPU, which is not optimized for neural networks. This integration makes it possible to implement more sophisticated algorithms, opening up new use cases.”
We are working with a number of partners to enable even more capabilities that take advantage of the performance and efficiency of our multimedia SoC processing platform for consumer IoT. Watch this space for more.
Vineet Ganju
Vice President of Marketing, IoT, Synaptics
