Promoted at CES 2025 and demonstrated as "a technological breakthrough," the new solution combines the use of Ambiq's latest Apollo510 MCU and neuralSPOT AI Deep Learning development platform to enable AI-powered consumer hearables and OTC hearing aids. Developers can use advanced AI-audio processing to create responsive and real-time hearable applications that feature enhanced voice with always-on listening.
According to Ambiq, the advanced capabilities of the Apollo510 ultra-low power system on chip (SoC) provide best-in-class performance and energy efficiency for all-day, practical use, while neuralSPOT accelerates the development cycle of manufacturers training, testing, and deploying audio AI models on their products. The solution presented at CES 2025 demonstrates the possibilities for noise reduction and sound enhancement in OTC hearing aids, with automatic adjustments picked up by the device's microphones.
With the sounds captured by these microphones, the Ambiq solution is able to differentiate between background noise, voices, traffic, music, and more, while the algorithms analyze, amplify, or reduce the different sound categories in real time. Ambiq believes this solution powered by its Apollo510 SoC addresses long-standing challenges for hearing-assist devices, through advanced neural network real-time processing.
Already available as part of the development kit is a new Neural Network Speech De-Noiser (NNSD) using an advanced deep learning model that analyzes audio in real-time to isolate speech and remove distracting background noise to improve user experience. This is combined with Neural Network Acoustic Environment Detection (NNAED), a highly sensitive deep-learning AI model that identifies subtle acoustic cues to adjust audio processing based on the user's environment dynamically. Finally, the solution uses advanced adaptive audio software that leverages AI to process the resulting experience with personalized sound settings adjusted by the user's habits and preferences.
"The acoustic scene detector is particularly powerful as it tunes the slight nuances in audio that often exceed the human ear's ability to distinguish between acoustic environments. Ambiq's extremely power-efficient compute capabilities made these AI features possible, representing a new milestone in AI sophistication for hearing aids," the company says.
The Apollo510 SoC pairs vector acceleration and power efficiency to enable most AI inferencing on-device without a dedicated NPU, offering the ability to do AI/ML workloads concurrent with complex graphics, telco-quality voice applications, and always-on voice/sensor processing. The Apollo510 MCU is a complete overhaul of hardware and software that fully leverages the Arm Cortex-M55 CPU with Arm Helium to reach processing speeds up to 250MHz. The Apollo510 achieves up to 10x better latency while reducing energy consumption by around 2x, compared to Ambiq’s previous Apollo4.
"By radically reducing the power needs of the highly complex audio processing and AI features required by advanced hearing aids, Ambiq has eliminated the need for specialized and expensive ASICs, accelerating product cycles while reducing costs," adds Carlos Morales, Vice President of AI at Ambiq. "Using our cutting-edge AI technology to improve sound quality, we can make OTC hearing aids more accessible and practical."
Although the number of OTC hearing aids and related products has expanded rapidly since the FDA's 2022 approval, a high rate of consumer returns has underscored the need for advanced audio technologies – a gap Ambiq aims to fill with advanced semiconductor technology. OTC hearing aids based on these sophisticated AI models allow Ambiq to offer a more energy-efficient solution to help manufacturers respond to the new challenging requirements.
www.ambiq.com | www.ambiq.ai
