Главная
Study mode:
on
1
AnalogML: Analog Inferencing for System-Level Power Efficiency
2
Today's Sensor Processing at the Edge is inefficient
3
Shifting the ML Workload to Analog
4
Efficiency with Analog
5
AnalogMLTM: Configurable Computing Chip
6
Analog Neural Network
7
Example of a Simple AnalogMLT Audio Chain
8
Application: Glass Break Detection
9
Application: Voice Activity Detection
10
Application: VAD + Preroll for WWE
11
Conclusion
Description:
Explore the innovative approach of analog machine learning (analogML) for ultra-low power audio event detection in always-listening edge applications. This tinyML Summit 2022 talk by David Graham, Co-founder and Chief Science Officer of Aspinity Inc., delves into the power inefficiencies of traditional sensor processing methods and introduces a solution that performs inferencing on raw, analog sensor data before digitization. Learn how analogML utilizes a library of software-configurable analog circuits programmed with standard machine learning techniques to enable highly discriminating event detection while significantly reducing power consumption. Discover the potential applications of this technology in voice-first devices, acoustic-based security systems, and other always-listening edge devices, and understand how it can dramatically improve battery life. Gain insights into the AnalogMLTM configurable computing chip, analog neural networks, and practical applications such as glass break detection, voice activity detection, and wake word engines. Read more

AnalogML - Analog Inferencing for System-Level Power Efficiency

tinyML
Add to list
0:00 / 0:00