New Method of Real-time Active Noise Reduction with AI

TapTechNews May 18th news, a research team from the University of Washington has recently invented a method to achieve real-time active noise reduction using AI, which can eliminate the sound specified by the user without changing the audio of the headphones.

The team led by artificial intelligence expert Shyam Gollakota presented this idea at a meeting held by the Acoustical Society of America and the Canadian Acoustical Association on May 16th and demonstrated a working prototype.

Gollakota and his team used a smartphone-based neural network to identify, train and filter 20 different environmental sound categories, such as the alarm and alarm clock sounds that people usually hear every day.

TapTechNews translated Gollakota's demonstration introduction as follows:

Imagine that you are enjoying the chirping of birds in the park, but the noisy chattering of a group of people next to you that you can't get them to stop talking.

At this time, when you put on headphones, it can allow you to focus on the chirping of birds and make other noises disappear. This is exactly the goal our system wants to achieve.

The prototype shows that there are microphones attached to both sides of the headphone ear cup, and the microphones are connected to an Orange Pi circuit board via USB, and the circuit board also provides audio to the headphones via the audio jack.

According to the layout of the circuit board, this is most likely OrangePi 5B, which uses the Rockchip RK3588S SoC (an octa-core 64-bit processor with a built-in neural processor with a computing power of 6 TOPS and can perform real-time filtering).

The team said that after AI training, it can complete active noise reduction in real time in less than one percent of a second without suppressing the actual sound played by the headphones.

Reference