Presentation Information

9:00 AM - 9:15 AM JST(12:00 AM - 12:15 AM UTC)Candidate for Academic Incentive AwardCareer Explorer

[A-14-13]ESP32 CSI Vowel Recognition System with Full-Subcarrier Statistics

◎△Toshihiro Kazama¹, Yosuke Sugiura², Tetsuya Shimamura¹ (1. Saitama Univ., 2. Meiji Univ.)

Keywords:

Speech Recognition,Wi-Fi Sensing,Embedded Wireless System,Device-Free Sensing,Channel State Information,CSI

Wi-Fi Channel State Information (CSI) can be used to observe subtle movements without wearing sensors on the body. While its effectiveness for lip and mouth movements has been demonstrated in systems such as WiHear [1] and RF lip-reading under masks [2], these methods sometimes rely on expensive wireless devices or limited experimental conditions. In this paper, we construct a system that identifies five vowels (a, i, u, e, o) based on mouth shape during speech using CSI acquired solely with a low-cost, readily available ESP32. Through this, we verify whether vowel recognition based on speech mouth shape is possible using a simple configuration that does not rely on specialized, expensive wireless devices.

Back to Session information