Telink Semiconductor unveils ML and AI TL-EdgeAI Platform

Artificial Intelligence(AI) serves as a vital driving force behind the current wave of scientific and technological revolutions, along with advancements in industrial processes. Presently, the deployment of AI applications has reached a new stage of development. The market that integrates wireless connectivity with edge AI computing capabilities is anticipated to experience significant growth in the near future, with AI applications at the edge representing a prominent aspect of this evolution.

Telink Semiconductor has recently achieved a new breakthrough in the chip technology field. It has officially launched the new-generation highly integrated chips TL721X and TL751X, as well as the Machine Learning (ML) and AI platform (TL-EdgeAI-DK) based on these two chips.

The TL721x and TL751x series have introduced advanced edge AI computing capabilities through the TL-EdgeAI platform. This platform is designed to support mainstream local AI models, including popular open-source frameworks such as Google LiteRT and TVM. It is recognized as the most energy-efficient smart IoT connection protocol platform available globally, thereby making it especially suitable for a variety of battery-powered applications. This development heralds significant opportunities for the expansion of extensive AI edge-side applications in the future.

As the industry’s first ultra-low-power multi-protocol IoT wireless SoC with an operating current as low as 1mA, TL721X is equipped to meet the rigorous standards imposed by the next generation of high-performance smart IoT terminal products.

Combining high performance, multi-protocol support, and a high degree of integration, TL751X utilizes an advanced multi-core design and incorporates a HiFi-5 DSP, resulting in exceptional computing and processing capabilities. Furthermore, this chip integrates a comprehensive range of practical functional modules, making it applicable across domains such as smart wireless IoT and wireless audio SoCs.

These two AI chips, along with the ML and AI platforms they encompass, enable Bluetooth chips to transcend their traditional roles of merely facilitating transmission. They are now capable of autonomous learning, engaging with large models, and implementing smaller models. Telink Semiconductor has also established extensive collaborations with leading manufacturers in the smart audio and smart home sectors, both internationally and domestically.

From Bluetooth Connectivity to Edge-Side AI: Telink Semiconductor Enters a New Development Stage

Since its establishment in 2010, Telink Semiconductor has been focusing on cutting-edge fields such as wireless IoT chip technology. It has a complete range of wireless IoT system-level chip products, including multi-mode IoT chips, wireless audio chips, and private protocol chips, meeting the diverse needs of IoT applications. For these chips, Telink Semiconductor provides self-developed firmware protocol stacks and complete reference designs, thus forming its product competitiveness and moat. The comprehensive products and technologies, the reliability verified by the market, and the stability of the supply chain have helped Telink Semiconductor rank among the top global connectivity chip manufacturers. By the middle of 2024, Telink Semiconductor’s global cumulative shipments had exceeded the significant milestone of 2 billion units.

Edge AI refers to the integration of artificial intelligence algorithms and processing capabilities on local or edge devices where data is generated, enabling devices to directly run AI models without relying on cloud servers for data processing. These local devices include a vast number of sensors, cameras, audio devices, robots, and other smart hardware. Through this technology, devices can process information in real-time at the location where data is generated, significantly reducing latency, saving data transmission bandwidth, and improving response speed. At the same time, since data does not need to be transmitted to remote servers for processing, it can effectively reduce the risk of data leakage and privacy issues.

The application scenarios of Edge AI are extensive, covering many fields such as smart homes, smart healthcare, intelligent buildings, industrial IoT, and security monitoring. Among the hundreds of millions of chips sold by Telink Semiconductor each year, a large number of products have been applied in these fields. In recent years, with the increasing demand from customers to add edge AI capabilities to chips, Telink Semiconductor has invested a large amount of R & D resources and finally launched a new generation of chips supporting edge AI computing. This breakthrough will significantly enhance the company’s product competitiveness in related fields and further open up the huge and rapidly growing market that requires both wireless connectivity and edge AI computing capabilities.

On the other hand, edge computing usually places high demands on device power consumption. Telink Semiconductor has a deep technical accumulation in the field of low power consumption, enabling the TL-EdgeAI platform to achieve the lowest power consumption performance in the world upon its release.

Efficiently Integrating Major Mainstream Algorithm Models to Open Up the Edge-Side AI Market

Based on the two highly integrated and ultra-low-power chips TL721x and TL751x, Telink Semiconductor has successfully created the TL-EdgeAI platform. This platform fully supports users in quickly porting existing machine learning models. Users can also use mainstream local artificial intelligence algorithms such as LiteRT and TVM, enabling them to efficiently integrate local artificial intelligence functions when developing actual products. Users can not only obtain directly runnable LiteRT models for flexible application in various machine learning and artificial intelligence scenarios but also use AI Edge conversion and optimization tools to smoothly convert TensorFlow, PyTorch, and JAX models into the TFLite format for operation.

Among them, TensorFlow and JAX models are developed and maintained by Google’s artificial intelligence team, Google Brain, while PyTorch is developed by Facebook’s artificial intelligence research team.

In terms of the workflow, it has also been greatly simplified. Taking LiteRT machine learning as an example, Telink Semiconductor’s local artificial intelligence platform already supports.

Users can use the ML/AI SDK provided by Telink Semiconductor and combine the results of the above – trained models to quickly integrate them into actual product applications and gain a head start in the market (Time-To-Market). Currently, Telink Semiconductor is actively assisting customers in introducing artificial intelligence functions using this platform, greatly enhancing product differentiation and creating product value. Telink Semiconductor has successfully integrated edge AI machine learning models into smart audio and smart home products using the TL-EdgeAI platform, achieving a close combination with practical applications. It is also collaborating with more users and strategic partners to develop innovative products with edge AI functions suitable for different application fields, including in-depth cooperation with leading domestic large-model AI companies on related products.

Take smart audio as an example. In consumer audio devices, TL751X supports high – quality audio transmission in devices such as headphones with its high performance, multi-protocol support, and high integration. At the same time, it uses its powerful computing power to assist in intelligent voice interaction. For example, it can intelligently switch audio modes according to voice commands. In audio conferencing systems, TL751X processes multiple channels of audio, achieving high-definition calls, audio mixing, and intelligent voice command response through artificial intelligence, such as automatically recording key points. In a multi-person networked intercom system, the TL-EdgeAI_751X platform, combined with self-developed low-latency Mesh network technology and Telink Edge AI-Noise Reduction technology, has achieved duplex communication with low latency, high stability, and high sound quality for up to 24 people simultaneously, promoting the continuous innovation and development of smart audio technology with the support of artificial intelligence and its wide application in various fields.

In the case of smart homes, the high performance, low latency, and low power consumption characteristics of TL721X make smart voice assistants more smooth and natural. Users can easily control devices such as lights, curtains, and air conditioners through voice with very low latency. Its support for Matter enables seamless connection of devices such as lights, curtains, and air conditioners with different protocols. The low power consumption advantage ensures the long – term operation of devices. Abundant peripheral interfaces can be connected to sensors. With the help of artificial intelligence algorithms, it can achieve intelligent perception, learning, and regulation of the indoor environment, greatly improving living comfort. Currently, the TL-EdgeAI_721x platform can support a variety of machine learning algorithms for smart sensors (Edge AI Sensor Hub) and the latest Matter network protocol. Telink Semiconductor is in a leading position in the expansion of smart home applications.

The launch of the new-generation highly integrated chips TL721X and TL751X and the TL-EdgeAI platform based on these two chips marks a significant progress for Telink Semiconductor in integrating machine learning and artificial intelligence software and hardware technologies and expanding the AI edge-side application market.

Currently, TL721X has entered the crucial stage of mass production preparation. It is expected to officially start large-scale mass production in the middle of 2025 and has already provided samples to some leading customers for early trial and evaluation. With its excellent high performance, wide multi-protocol support, and outstanding high integration, TL751X has already made a name for itself in the market.

Telink Semiconductor will integrate edge AI machine learning models into smart audio and smart home products through the TL-EdgeAI platform, achieving a close combination with practical applications. It is also collaborating with more users and strategic partners to develop innovative products with edge AI functions suitable for different application fields. The launch of the TL-EdgeAI platform will enhance the competitiveness of Telink Semiconductor’s products in related fields and further open up the huge and rapidly growing market that requires both wireless connectivity and edge AI computing capabilities.