Comparative Study of Digital Filters for a Smart Glove Functionalized with Nanocomposite Strain Sensor

Nowadays, the demand for flexible and wearable measurement systems is significantly increasing. Thereby, sensors based on carbon nanotubes (CNT) materials are gaining importance due to their high flexibility, sensitivity and medical compatibility. Nanocomposite based strain sensors present an interesting potential of movements detection that helps to build the basis for body attached sensor networks, which have the capability for tracking finger movements, gestures and grasping. However, the output of the sensor is noisy and having outliers due to the random deformation of the CNT conductive paths during applied strain, which needs to be filtered. This paper presents S-TPU/C-TPU strain sensors attached to a glove for hand gesture recognition. A filtering study has been conducted of four digital filters for real-time processing of the recorded data (Moving average, Moving median, Savitzky-Golay, and Gaussian), which is required to improve the quality of the data for better gesture identification and classification. The filters are evaluated based on their real-time response and smoothing performance. The Savitzky-Golay filter was found to be the most suitable filter to remove the inherent noises and smooth the baseline and curves of the signal with a fast real-time response equal to 0.072s.