Sitting Posture and Identity Recognition of Flexible Capacitive Sensors Integrated With Deep Learning Models

ABSTRACT With the growing demand for health monitoring and intelligent interaction, developing low‐cost, high‐performance flexible pressure sensors has become a research hotspot in human‐machine interaction. This study proposes a simple, low‐cost method to fabricate 7 flexible capacitive pressure sensors for efficient sitting posture detection and identity distinction. The sensor has a three‐layer structure: a copper foil upper electrode, a polyimide (PI) tape dielectric layer, and a lower electrode (PPA@MS) composed of melamine sponge impregnated with transparent conductive ink (main components: PEDOT/PSS and AgNWs). It exhibits excellent performance: high sensitivity (4.86% kPa − 1 at 0–20 kPa), low hysteresis (7.08%), fast response/recovery time (100 ms/120 ms), and good stability over 5000 cycles. Mounted on a seat, the 7 sensors collect sitting posture signals. A Convolutional Neural Network‐Long Short‐Term Memory (CNN‐LSTM) deep learning model with superior feature fusion capability is built, achieving 95.68% accuracy for 9 postures and 95.19% for 9 identities. Results show the sensor can efficiently capture and distinguish posture signals; the CNN‐LSTM model has broad prospects in human‐machine interaction, and the sensor‐model combination is expected to be widely used in health monitoring and smart homes.