Decoding of Turning Intention during Walking Based on EEG Biomarkers

Abstract

In the EEG literature, there is a lack of asynchronous intention models that realistically propose interfaces for applications that must operate in real time. In this work, a novel BMI approach to detect in real time the intention to turn is proposed. For this purpose, an offline, pseudo-online and online analysis is presented to validate the EEG as a biomarker for the intention to turn. This article presents a methodology for the creation of a BMI that could differentiate two classes: monotonous walk and intention to turn. A comparison of some of the most popular algorithms in the literature is conducted. To filter the signal, two relevant algorithms are used: H∞ filter and ASR. For processing and classification, the mean of the covariance matrices in the Riemannian space was calculated and then, with various classifiers of different types, the distance of the test samples to each class in the Riemannian space was estimated. This dispenses with power-based models and the necessary baseline correction, which is a problem in realistic scenarios. In the cross-validation for a generic selection (valid for any subject) and a personalized one, the results were, on average, 66.2% and 69.6% with the best filter H∞. For the pseudo-online, the custom configuration for each subject was an average of 40.2% TP and 9.3 FP/min; the best subject obtained 43.9% TP and 2.9 FP/min. In the final validation test, this subject obtained 2.5 FP/min and an accuracy rate of 71.43%, and the turn anticipation was 0.21 s on average.

Keywords: ASR; BMI; EEG; H∞; intention turn direction; real time.

Figure 1

In the image on the left, the subject with the equipment on the day of the test. With this setup, the training and test were performed. The training and test protocol is schematized in the figure on the right.

Figure 2

On the left: Image of the processed IMU signal. In black, the moment classified as turn. In blue highlighted, the average number of turns on the left and in red highlighted, the average number of turns on the right. In green, the turns that did not meet the criteria assigned to the turn type (left or right) assigned by the algorithm. On the right: the EEG signal of the third turn for the Fz channel of subject S2. Labeled with respect to the turning point marked on the IMUs. For the segmented class in blue monotonous walk and for the intention turn class in red. According to the raw EEG signal (first image), the signal filtered with $H^{\infty }$ (second image) and the signal filtered with ASR (third image).

Figure 3

Pseudo-online with 16 repetitions. Horizontal line at 0, in yellow the mean of the first moment in which the PFs are evaluated. In blue, the temporary maximum where it is evaluated in yellow the mean of the time repetitions duration. The FPs are shown in red as they happened temporarily with respect to the change in the X axis, the value of Y is based on the percentage of times that they appeared at that moment. The red horizontal line is the average number of trials in which an FP appeared. Analogously, the TPs are shown in green and the average PT horizontally.

Figure 4

The 2D dimensionality Riemannian reduction in the Riemannian tangent space of eight trials chosen to make the model of each subject. The bands chosen for each model are those of the personalized configuration $H^{\infty }$. In blue, the epochs of monotonous walk and in orange the epochs of intention to change direction.

Figure 5

Color map of the upper diagonal of the matrix of differences between the mean of the matrices of the monotonous walk and change intention classes for each of the subjects in the best configurations in Table 4. The best electrodes for the selection of 15, 10 and 5 are indicated respectively underlined with colors: green, yellow and red.

Figure 6

Figure with the FP of the subject who performed the test in real time. On the Y-axis, the number of repetitions performed and on the X-axis, the time elapsed since the start of the task that supports the sending of BMI commands. If the BMI detects the class intention to turn, the instant (frame) is displayed in yellow. If four consecutive FPs occurred, the person was forced to turn (instant of the verbal command marked with an X) and in the figure the remaining moments are plotted in red.