Lower leg muscle activity contributes to body control; thus, monitoring lower leg muscle activity is beneficial to understand the body condition and prevent accidents such as falls. Amplitude features such as the mean absolute values of electromyography (EMG) are used widely for monitoring muscle activity. Garment-type EMG measurement systems use electrodes and they enable us to monitor muscle activity in daily life without any specific knowledge and the installation for electrode placement. However, garment-type measurement systems require a high compression area around the electrodes to prevent electrode displacement. This makes it difficult for users to wear such measurement systems. A less restraining wearable system, wherein the electrodes are placed around the ankle, is realized for target muscles widely distributed around the shank. The signals obtained from around the ankle are propagated biosignals from several muscles, and are referred to as distal EMG signals. Our objective is to develop a sock-type wearable sensor for estimating lower leg muscle activity using distal EMG signals. We propose a signal processing method based on multiple bandpass filters from the perspectives of noise separation and feature augmentation. We conducted an experiment for designing the hardware configuration, and three other experiments for evaluating the estimation accuracy and dependability of muscle activity analysis. Compared to the baseline based on a 20-500 Hz bandpass filter, the results indicated that the proposed system estimates muscle activity with higher accuracy. Experimental results suggest that lower leg muscle activity can be estimated using distal EMG signals.

Surface EMG (sEMG) signals are useful for estimating the motion or exercise of users. Wireless-type sensor electrodes, which are placed on multiple parts of the body and send the measured signals to a server, have recently become commercially available. With many estimation algorithms, the relationships between the sensor IDs and the body parts they are placed on (ID configuration) are expected to be fixed between the calibration and estimation phases. If the ID configuration is changed after the calibration phase, the estimation accuracy tends to dramatically decrease. Since it is inconvenient for users to check the ID configuration every time, we developed a method to correct the electrode ID configuration on the basis of the distribution of sEMG features. Using open data, we investigated the feasibility of our method by shuffling the order of sEMG signals. The results showed that the method was able to correct the ID configuration and restore the estimation accuracy to close to that of the calibration.