EEG motor imagery classification using deep learning approaches in naïve BCI users.

Motor Imagery (MI)-Brain Computer-Interfaces (BCI) illiteracy defines that not all subjects can achieve a good performance in MI-BCI systems due to different factors related to the fatigue, substance consumption, concentration, and experience in the use. To reduce the effects of lack of experience in the use of BCI systems (naïve users), this paper presents the implementation of three Deep Learning (DL) methods with the hypothesis that the performance of BCI systems could be improved compared with baseline methods in the evaluation of naïve BCI users. The methods proposed here are based on Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM)/Bidirectional Long Short-Term Memory (BiLSTM), and a combination of CNN and LSTM used for upper limb MI signal discrimination on a dataset of 25 naïve BCI users. The results were compared with three widely used baseline methods based on the Common Spatial Pattern (CSP), Filter Bank Common Spatial Pattern (FBCSP), and Filter Bank Common Spatial-Spectral Pattern (FBCSSP), in different temporal window configurations. As results, the LSTM-BiLSTM-based approach presented the best performance, according to the evaluation metrics of Accuracy, F-score, Recall, Specificity, Precision, and ITR, with a mean performance of 80% (maximum 95%) and ITR of 10 bits/min using a temporal window of 1.5 s. The DL Methods represent a significant increase of 32% compared with the baseline methods (p< 0.05). Thus, with the outcomes of this study, it is expected to increase the controllability, usability, and reliability of the use of robotic devices in naïve BCI users.

Semi-supervised approach to identify steps, shoes, and gender of older adults under semi-naturalistic conditions using a waist-worn inertial sensor.

An emerging source of information to recognize individuals' characteristics are the walking pattern-related parameters. The elderly can be one of the populations that can benefit most from recognition-based applications, which may help to increase their possibilities of living independently at home. Approaches have been mostly focused on gait events' identification or assessment; nonetheless, such information can also be used to obtain seniors' characteristics that depend on physiological or environmental factors. These factors can be useful to provide a customized assistance based on contextual information. In this paper, we propose a method focused on seniors, to detect steps, and to recognize gender and type of shoes by using only the initial foot contact (IC) data obtained from inertial sensors during semi-controlled walking. Data were collected from 20 older adults who walked at self-speed in a natural environment. The method consists of first clustering the IC using k-means; then, a trained recurrent neural network recognizes gender, type of shoes, and the step phases (IC and other phases); to finally conduct step detection (SD) using a ruled-based method. The method recognizes gender and the type of shoes with an accuracy of 93% and 83.07%, respectively, whereas there were not misrecognitions of the step phases. SD achieved a mean absolute percentage error equal to 0.64%. The good results show that the method is appropriate for users' characteristics recognition applications without depending on assumptions based on individualities. Likewise, the method can be useful to monitor physical activity or systems aimed to keep safe older adults.