Low-Crosstalk Dual-Channel Temperature–Pressure Sensor Array Based on PEDOT:PSS/MXene/Carbon Foam Composites for Multimodal Recognition

Reliable electronic skin requires thermal and mechanical inputs to be resolved in the same soft device, but stacked sensor architectures often add fabrication complexity and interfacial coupling. Here, we presented a low-crosstalk dual-channel temperature-pressure sensor based on PEDOT:PSS/MXene/carbonized melamine foam (PMCF) composites, in which pressure and temperature are transduced through distinct physical processes. Reversible contact reconstruction within PMCF defined a piezoresistive channel that modulated conductance under pressure, whereas the PEDOT:PSS/MXene network determined a thermoelectric channel that converted a temperature gradient into voltage. The integrated architecture enabled pressure-dominated conductance and temperature-dominated voltage readouts within the PMCF composites. The thermoelectric channel produced a linear voltage response within ΔT of 0-50 K, a Seebeck coefficient of 37.9 μV K-1, and stable output over 12,000 s of thermal cycling, whereas the piezoresistive channel presented a linear response over 1-20 kPa, with a sensitivity of 1.6409 ± 0.016 kPa-1 and stable output during load-unload cycles for 7000 s. The minimum temperature-difference resolution was 0.1 K. Temperature/pressure channels retained their dominant stimulus origins, with pressure-to-voltage and temperature-to-current crosstalk ratios below 0.40% and 0.35%, respectively. The PMCF sensor facilitated subtle contact acoustic-vibration sensing and array-level multimodal input. A CNN-BiLSTM-attention model from machine learning was employed to classify five acoustic-vibration signals with an accuracy of 97.9%. A CNN model, applied in a 3 × 3 sensor array, recognized digit position, input mode, and joint digit-mode patterns with accuracies of 97.4%, 99.1%, and 96.4%, respectively, indicating that the PMCF sensor array supported low-crosstalk multimodal sensing and array-level human-machine input under controlled conditions.