PVDF tactile sensor based on time-series image encoding

To enhance the recognition accuracy and real-time performance of flexible tactile sensing systems in human–computer interaction, this paper proposes a tactile recognition method based on time-series imaging with polyvinylidene fluoride (PVDF) sensors. A 4 × 4 PVDF array is constructed, and the raw tactile signals are acquired and preprocessed before being transformed into Gramian angular summation field (GASF) and Markov transition field (MTF) images. These images are then classified and analyzed by a convolutional neural network (CNN) for contact state recognition and trajectory reconstruction. The experimental results show that the four classification metrics of both methods are above 95%, and the recall and F1-score of MTF+CNN are slightly higher than those of GASF+CNN. Under the hardware environment of a MacBook Pro 2022 (M2 chip with GPU acceleration), the average inference latency per sample is maintained at the millisecond level, demonstrating the effectiveness of the proposed method for tactile recognition.