So far, there has still been a challenge to eliminate the signal interference in the decoupling analysis of flexible temperature sensors. Herein, a solution-based route was employed for doping thermoplastic polyurethane (TPU), carbon black (CB) and ionic liquid TFSI into polyvinylidene fluoride (PVDF), leading to a flexible, conductive composite membrane with both temperature and pressure sensing characteristics. Such a TPU/PVDF/TFSI@CB sensor demonstrated an excellent sensitivity of 1.24 %/℃ over the wide range of 55–85 ◦ C. Electrical con ductivity is enhanced through TFSI doping, and interfacial stability is optimized. Moreover, stretchability is boosted to 375 % and temperature response is enhanced by 85.41 % at 85 ◦ C with TPU incorporation. After wards, the sensors were successfully applied to human health and motion monitoring, including finger bending, blowing, and throat swallowing. Further, the interference from pressure could be eliminated effectively through using a facile machine learning, enabling a temperature recognition accuracy of 98 % via PCA. We believe this smart recognition framework can be generalized to other temperature sensors as well.