Deep Learning Based Non-Contact Communication using Air-Coupled Leaky Lamb Wave

Abstract

Ultrasound is used as an alternative to conventional communication in scenarios with severe electromagnetic shielding and interference. Ultrasonic guided waves, like Lamb waves, exhibit favorable transmission characteristics in platy structures, enabling long-distance propagation with low energy loss. This paper proposes a non-contact data transmission method using air-coupled leaky Lamb waves. Wave conversion and multipath interference have nonlinear effects on air-coupled leaky Lamb wave communication. To decode received temporal signals, a lightweight deep residual network (ResNet) designed for end-to-end time series classification (TSC) is employed. Real-world transmission experiments on a platy structure with a pitch-catch configuration were conducted, comparing three modulation methods: amplitude shift keying (ASK), frequency shift keying (FSK), and binary phase shift keying (BPSK). The trained ResNet achieved 1 recognition accuracy for each modulation. In bit error rate (BER) experiments, the error rate for each modulation was 0.