Microwave Speech Recognizer Empowered by a Programmable Metasurface.

Area of Science:

• Microwave Engineering
• Artificial Intelligence
• Human-Machine Interaction

Background:

• Speech recognition is crucial for human-machine interaction but struggles in real-world conditions due to noise and obstructions.
• Existing systems are limited by acoustic signal degradation and visual occlusions (e.g., face masks, walls).

Purpose of the Study:

• To develop a privacy-respecting microwave speech recognizer capable of operating in challenging environments.
• To demonstrate the efficacy of programmable metasurfaces in enhancing speech recognition capabilities.

Main Methods:

• An experimental prototype utilizing a programmable metasurface and artificial intelligence was developed.
• The system processes microwave data, bypassing the need for optical sensors and acoustic signals.
• The programmable metasurface's large aperture and degrees of freedom enable complex sensing tasks.

Main Results:

• The prototype successfully recognized voice commands and speaker identities remotely, even with occluded mouths or in noisy settings.
• The system demonstrated privacy-respecting human-machine interaction by relying solely on microwave data.
• Experimental validation confirmed performance in previously inaccessible application scenarios.

Conclusions:

• The programmable metasurface-empowered microwave speech recognizer overcomes significant limitations of current systems.
• This technology offers potential for advanced applications in smart homes, health monitoring, and security.
• The approach enables robust and private voice-controlled interactions.