Cupertino, CA, September 30, 2024 – Aizip, Inc., a leading technology company specializing in artificial intelligence (AI) applications for endpoint and edge devices, today announced the expansion of its Audio Enhancement Model portfolio to address a broader range of applications.
AI has revolutionized audio processing, delivering better performance, new applications, and making it better than DSP-based audio technology for many applications. AI-based audio offers two major advantages. First, AI models have stronger non-linear modeling capabilities than traditional hand-crafted algorithms, which allows for more-accurate signal separation and superior overall performance. Second, these models are trained using vast data across a variety of scenarios, providing enhanced adaptability by eliminating the need for detailed parameter-tuning in different environments.
Expanding on Aizip’s industry-leading ultra-efficient Deep Noise Reduction (DNR) technology, the Aizip audio team now offers three new ultra-efficient solutions: Deep Noise Reduction (DNR) with model latencies of 4ms, real-time Music Source Separation (MSS), and AI-based Acoustic Echo Cancellation (AEC). These solutions significantly broaden Aizip’s product portfolio into a comprehensive audio-enhancement suite.
- Aizip had previously developed an ultra-efficient DNR solution with latency as low as 8ms. To achieve even lower latencies, however, the team developed the new Zenet picoMini architecture that reduces computational requirements to enable real-time performance on tiny devices with only 4ms latency. This breakthrough architecture enables higher-quality AI audio processing even on devices with the most stringent requirements, such as hearing aids and same-room voice amplification.
- A growing application in the portable wireless-speaker market and automobile markets is karaoke. Aizip has developed a real-time AI-based vocal-track removal model called Karaoke-MSS that can be deployed on the microcontroller in a variety of devices, including the speaker itself, to stream karaoke anytime, anywhere.
- Acoustic Echo Cancellation (AEC) is a classic challenge in audio processing, where an interference source needs to be removed using a reference signal. Traditional AEC solutions involve two stages: Linear AEC and Non-Linear Residual Echo Cancellation (REC). Conventional REC can degrade voice quality. Aizip’s AI-based REC improves performance without adding latency while also incorporating noise reduction to remove non-speech interference.
Aizip’s head of business development Nathan Francis is looking forward to these innovations. “These audio solutions significantly enhance application performance and are fun for people around the world. We expect products enabled by our AI to hit the market by the holiday season this year.”
Dr. Boltzmann Li, head of Aizip’s Intelligent Audio (AIA) product development, emphasized the company’s commitment to continued innovation. “AI will play a transformative role in audio technology. We’re working closely with our partners to deliver products that stand out in the market. And we are developing many more applications that will enrich people’s audio experience globally.”
Aizip provides TinyML models to IC partners, module providers, and system integrators worldwide. The team is dedicated to delivering low-cost, high-performance AI solutions that benefit society. For more information, contact info@aizip.ai.
* * *
About Aizip, Inc.
Based in the heart of Silicon Valley, Aizip, Inc. develops advanced AI models for endpoint and edge device applications. Renowned for its model performance, fast deployment, and superior return on investment, Aizip provides versatile solutions for a wide range of intelligent, automated, and interconnected applications. Learn more at www.aizip.ai.