A new neuromorphic chip has shown it can perform continuous voice activity detection while using extremely low power, addressing a major challenge in devices that rely on batteries or operate at the edge of a network. The chip identifies human speech with very fast response times by running neural network calculations in the analog domain. This method keeps voice detection running without using much energy, unlike traditional digital AI systems that use a lot of power constantly or depend on periodic checks to wake up.
Voice Activity Detection is a common need for many devices, such as wearables, smart sensors, robots, and communication tools. Keeping microphones and processors running all the time usually uses up too much power for systems that need to save energy. The neuromorphic design helps save both power and time by processing signals in the analog domain.
The current demonstration focuses on detecting voices, but the same technique can be used for more complex voice-related tasks like identifying speakers or separating voices. These features could allow edge devices to process speech locally, reducing the need for cloud-based processing. This progress demonstrates how analog neuromorphic computing is moving from being studied in labs to being built into real hardware, especially for applications that need to use energy wisely and respond quickly.
