While smart glasses such as Meta’s Ray-Bans are increasingly seen as ‘pervert glasses’, the technology is powerful in assisting people with hearing, sight and memory issues. The video camera, in particular, is seen as an intrusion into people’s privacy, allowing wearers to film people without their knowledge or consent. So there are alternative approaches for using the assistive technology. These include glasses with open-ear audio amplification, transcribing the audio, and even lip reading from underneath a mask. New physics-based simulation tools for AI are enhancing the development of these glasses, while a new generation of ultra-low power processors with ARM’s Cortex-M85 core is helping to improve the performance, host local AI and increase the battery life. “I’ve been really concerned by the recent misuse of AI/Smart glasses,” said Richard Boggie , a blind playwright in Newcastle, UK. “I am also appalled at the way some people are using the technology. It’s exploitative and inappropriate, but I’m worried that the public reaction to this misuse will lead to genuine users being challenged.” More than 1.2 million adults in the UK have hearing loss severe enough to make ordinary conversation difficult, according to the Royal National Institute for Deaf People, and video is not necessary to help them. Smart glasses developed by Nuance Audio use an array of six internal microphones combined with directional beamforming technology and open-ear speakers to amplify speech without putting anything inside the ear. Enhanced directional audio with smart glasses from Nuance Audio Six microphones are hidden along the frame: two facing forward at the front and others tucked along the temples. They use beamforming algorithms to lock onto sound coming from a specific direction while reducing competing background noise from the sides and rear. Discreet mini-speakers positioned in the temple arms just above the ears deliver the processed sound cleanly via air conduction. A low latency of under 8 milliseconds ensures that lip movements and the amplified sound stay synchronized. There are several listening modes. The Frontal Mode narrows the focus onto the person directly ahead and trims down ambient noise from the sides and back for environments such as a restaurant or one-on-one conversation, while the All-Around Mode opens up the audio capture evenly across 360 degrees for walking, driving or general awareness. Other developers such as Even Realities in Shenzhen, China, are also avoiding using cameras. Instead, the Even G2 smart glasses display transcripts of conversations on a screen. This AI-enabled ‘conversate’ feature offers prompts and explanations during conversations, and generates AI summaries afterwards. The 36g smart glasses also provide translations of more than 30 languages and navigate with a built-in geomagnetic sensor in the Even R1 smart ring. AI modelling Modelling how the sound is heard is a key challenge, especially for AI models. Treble in Iceland is developing simulation tools to model the performance of AI models, and smart glasses are a rapidly growing areas, says co-founder Finnur Pind. “We are modelling the physics of sound, to create the acoustics of any conceivable environment, “he said. “We work with models of microphones, loudspeakers, analog-to-digital converter (ADC), digital-to-analog converter (DAC) and the head response transfer function (HRTF) and model the physics of all those components, for example a loudspeaker in AR glasses, we model the physics of the loudspeaker and the input radiates around the room, reaches the glasses, we model the physics of the microphone refraction around the head.” Modelling tools for AI smart glasses “Smart glasses is one of our bigger markets at the moment, modelling the shape of the head and the fitting of the glasses,” said Pind. “Our customers are the ones developing the LLM and they do both edge and remote processing.” “I’m very impressed by what can be done today on d