Cupertino, CA, March 30, 2024 – Aizip, Inc., a vanguard in artificial intelligence (AI) for endpoint devices, known as TinyML, is proud to announce today its achievement of several milestones in its cutting-edge AI-based Deep-Noise-Reduction (DNR) models, designed for a variety of voice applications.
Noise reduction stands as a critical challenge in the domain of audio processing. Leveraging AI’s transformative capabilities, Aizip has addressed complex acoustic disturbances such as wind noise and low signal-to-noise ratio scenarios with unmatched proficiency. Initiated shortly after Aizip’s inception in 2020, the DNR program’s cornerstone is the innovative Zenet neural network architecture. This technological marvel has demonstrated exceptional efficiency and effectiveness, catering primarily to the endpoint device market.
Aizip has successfully developed and dispatched TinyML models to an array of customers and partners, encompassing a wide breadth of applications. With a focus on product robustness, two models have been refined: a 260kB version for stringent quality demands and a 60kB variant for cost-effective solutions. These models have been employed in diverse contexts, including True Wireless Stereo (TWS), microphones, and wearable technologies. The company anticipates the commercial shipping of these customer products to commence in the forthcoming month.
In parallel to these business triumphs, the Aizip team has achieved significant technical advancements, further cementing their technological supremacy. A recent focus on minimizing latency—a crucial factor in automotive and gaming industries—has culminated in the release of a commercial DNR model boasting a 16ms latency. “This model has been subjected to rigorous testing, where it showcased outstanding performance and reliability,” remarks Matthew Kymn, an esteemed AI audio engineer at Aizip.
Another remarkable feat by Aizip involves reconstructing voice communication via bone conduction amidst extremely noisy environments. Utilizing bone conduction sensors, now commercially accessible from developers like Bosch and Knowles, Aizip has exhibited stellar noise reduction capabilities, especially in recovering voice signals missing high-frequency components. “Our technology ensures clear voice restoration even at SNR levels as low as -15dB,” says Boltzmann Li, the visionary leader behind these auditory breakthroughs. Dr. Li reveals that while the model is prepared for customer evaluations, the pursuit of perfection continues as improvements are actively being made.
The DNR technology is a key component of Aizip’s Intelligent Audio (AIA) portfolio, which has garnered widespread endorsement from partners and clientele. Committed to innovating the auditory experience, Aizip’s team is steadfast in pushing the boundaries of AI applications within the audio industry. Offering TinyML models globally, Aizip is dedicated to providing low-cost, high-performance solutions with the aspiration of benefiting society at large. For additional information, please reach out to info@aizip.ai.
* * *
About Aizip, Inc.
Situated in the heart of Silicon Valley, Aizip, Inc. specializes in developing superior AI models tailored for endpoint and edge device applications. Aizip stands out for its exemplary model performance, swift deployment, and remarkable return on investment. These models are versatile, catering to a spectrum of intelligent, automated, and interconnected solutions. Discover more at www.aizip.ai.