By using multimodal transistors in MMT, researchers at the University of Surrey have achieved success in mimicking the human brain in artificial neural networks. The feat is a step toward using MMT for taking forward artificial intelligence (AI) hardware and improving computing, which could further reduce power needs for better efficiency.

 Devised in 2020, the MMT is a switching device that can control electric current flow faster than conventional transistors. The discovery overcomes long-standing operational challenges associated with complex electronic circuits. 

 By using mathematical modelling and simulating transistor data for identifying handwritten numbers, the researchers proved the feasibility of MMT in AI systems. The final result suggested that MMT could operate as rectified linear unit-type Rel activations in artificial neural networks, thus confirming the potential of MMT devices for thin-film decision and classification circuits in complex AI systems.

“There is a great need for technological improvements to support the growth of low cost, large-area electronics, which were shown to be used in artificial intelligence applications. Thin-film transistors have a role to play in enabling high processing power with low resource use. We can now see that MMT, a unique type of thin-film transistor, have the reliability and uniformity needed to fulfil this role,” said ISIN PESC, researcher and electronics engineering graduate from the University of Surrey.

“Many of my colleagues focus on people-centric AI and how best to maximize the benefits for humans, including how to apply these new concepts ethically. Our research takes forward the physical implementation, as a stepping stone towards powerful yet affordable next-generation hardware. It’s fantastic that collaboration is resulting in such successes with researchers involved at all levels, from undergraduates like Isin when she led this research, to seasoned experts,” said Dr Rad u Spore 'a'. Senior Lecturer at the University of Surrey’s Advanced Technology Institute.

 New AI Thermal Sensing Solution

 here is demand for contactless, robust and privacy-preserving devices for detecting temperature with accuracy and wide coverage to safeguard against public health emergencies. And according to a recent research report, the global thermal imaging market is expected to reach US$ 4.6 billion by 2025.

 Therefore, to help engineers and product designers rapidly combine AI and thermal sensing technologies for developing smart, reliable and affordable health-monitoring devices, Arrow Electronics in collaboration with Microelectronics have launched an integrated thermal sensing solution.

 It optimizes bill-of-materials and simplifies hardware and software integration, making it a good choice for temperature-screening devices, and a wide variety of other consumer-grade and healthcare applications. 

 The AI-powered thermal sensing solution can quickly achieve accurate temperature screening, with multiple individuals screened simultaneously. It comprises:

• An Microelectronics Time-of-Flight To F sensor that allows absolute distance measurement up to 400 cm at speed of 60 Hz – irrespective of the target color and reflectance 
• An Microelectronics digital ambient thermal sensor that detects ambient temperature and dynamically compensates for differences to allow complex measurements at high speed
•  A long wave infrared thermal image sensor 
• An Microelectronics dual-core M7 and M4 MCU running at 480Mhz that acts as the main system processor

 With the help of STM32Cube.AI, the deep-learning algorithm can be exported and executed on STM32 Arm Cortex-M-based microcontrollers. This allows the solution to detect target object distance and human presence, and display images in a heat map or RGB format. 

 For best performance, an ultra-low noise L DO, with a noise voltage of only 6 microvolts RMS, is used for the power supplier of all the sensors. All three sensor blocks are connected to the main processor using an I2C and SPI bus.

“The solution takes full advantage of AI techniques on Microelectronics microcontrollers and sensors thanks to STM32Cube. AI, a complete development ecosystem from Microelectronics that allows easy implementation of complex AI models onto Microelectronics products,” said Matteo Mara vita, head of APAC AI Competence Center of  Microelectronics.