Accelerating AI-Enabled Content Requires New Levels of Complexity Management
Sep 13, 2022
By Abdel Younes
The introduction of artificial intelligence (AI) to multimedia processing has resulted in an explosion of new features, capabilities, and applications. Soundbars, set-top boxes (STBs), TVs, and digital signage have evolved into interactive, feature-rich devices with camera and microphone interfaces to take advantage of AI-enabled features.
In the home, these capabilities range from Dolby Atmos 5.1 and Dolby Vision, volume equalization, dialog enhancement, and video up-scaling to facial and speech recognition, tracking, and body pose estimation. In public spaces, AI has enabled people counting and gender and age identification for intelligent signage.
The continuing shift to AI at the Edge enables intelligent decisions to be made locally instead of in the cloud. This enhances content security and user data privacy while reducing latency and optimizing bandwidth usage.
But all of this introduces new challenges and complexities for system developers.
Synaptics is focused on delivering new types of AI-enriched solutions that can manage the complexity and combine the necessary elements. These elements include: high performance, low power, and secure hardware architectures; complex, application-specific AI algorithms; and open source, use case-specific training models that enrich the user experience.
Our VS680 multimedia processor delivers the throughput, power efficiency, and scalability required for a range of intelligent content delivery systems and applications. Scalability comes through the platform’s mix of high-performance CPU, GPU, and dedicated neural processing unit (NPU), and supporting peripherals. This highly integrated SoC’s security-first approach ensures that the NPU and the AI audio/video pipeline are executed in a secure environment, providing user privacy and ensuring that the system meets DRM requirements.
A growing ecosystem of developers, software, and system providers are optimizing models for this platform, using our Android-based, open-source SyNAPTM kit for an easy and highly secure path to AI model optimization. Pilot.AI, Visidon, Askey, and IDR&D, are already using the platform for custom deployments.
The kit’s features include the execution of an Executable Binary Graph (EBG)—a Synaptics-designed high-performance format—as well as the ability to sign and encrypt the EBG (SyKURE) for trusted and secure inference.
The AI-enabled content processing features of Synaptics’ VS680 SoC, SyNAP and SyKURE will be on display at IBC: email email@example.com to book time at our IBC Balcony Booth BS17.