Cupertino, CA, September 28, 2022 – Aizip, Inc., a technology leader in artificial intelligence (AI) for applications on the Internet of Things (IoT), also known as tiny machine learning (TinyML), announced today the release of robust audio models and the beginning of volume shipping.
Audio-based intelligence plays an important role in smart homes, smart cities, Industry 4.0 applications, and many other markets. Voice command-controlled home appliances liberate consumers from pushing buttons. Glass-breaking detection helps to protect the home or office from burglars. Speaker identification allows a music player to choose customized playlists. Audio TinyML models contribute significantly to enriching people’s lives.
Aizip began developing audio models soon after it was founded two years ago and has developed models to cover an extensive range of audio applications. The voice-command product line includes wake-words recognition, keyword spotting (KWS), spoken language understanding (SLU), automatic speech recognition (ASR), and speaker identification. The event-detection models cover baby crying and glass breaking and are being expanded to other sound events. AI-based deep noise reduction (NDR) has demonstrated excellent performance at a compact size.
In parallel to expanding audio applications, Aizip has invested significant resources to bring audio models from demo to production quality. “While it’s already challenging to build TinyML models for small IC devices, it’s ten times more challenging to bring in robustness for real-world applications,” said John Rea, Chief Operations Officer of Aizip. Although public datasets are great sources for demos and benchmarking, more real-life data is needed to cover a wide range of practical cases, according to Rea. He further shared that Aizip has worked closely with its partners and completed many months of extensive testing and model optimization to achieve strong robustness.
“Building on top of our breakthrough neural architectures last year, we’ve made significant further progresses in audio technology,” commented Boltzmann Li, who leads the audio-product team at Aizip. One example shared by Li is contextualized detection that dramatically improves audio-detection accuracy. These innovations help Aizip to build a strong IP portfolio and product differentiations. The resulting models perform very well in real-world testing and have been well received by Aizip customers, leading to volume orders and shipping.
The robustness methodology and techniques developed in Aizip Intelligent Audio (AIA) have also been applied to Aizip Intelligent Vision (AIV) and Aizip Intelligent Time-Series (AIT) products. Aizip provides TinyML models to its IC partners and module and system customers worldwide. Aizip is committed to developing low-cost and high-performance intelligence for the benefit of society worldwide. For more information, please contact info@aizip.ai.
* * *
About Aizip, Inc.
Aizip, Inc., develops AI models for edge applications. Based in Silicon Valley, Aizip provides models with superior performance, quick turnaround time, and excellent ROI. These models can be used in a wide range of applications for an intelligent, automated, and connected world. For more information, visit www.aizip.ai.