Presentation title
User-Centred BCI for Mechatronic Actuation by Spatio-Temporal P300 MonitoringAuthors
Daniela De Venuto, Giovanni Mezzina and Valerio Francesco AnneseInstitution(s)
Politecnico di Bari, ItalyPresentation type
Technical presentationAbstract
This work presents a P300-based Brain Computer Interface (BCI) for the remote control of a mechatronic actuator, such as wheelchair, or even a car, driven by EEG signals to be used, for instance, by tetraplegic and paralytic users. The P300 signal is an Evoked Related Potential (ERP) often used as a biomarker for the assessment of the cognitive brain activity. It is basically induced by visual stimulation protocol known as odd-ball paradigm. In our application a four-choices odd-ball paradigm has been implemented (e.g. T1, T2, T3, T4) The realized BCI is based on a machine learning algorithm, which exploits a spatio-temporal characterization of the P300, analyses all the binary discrimination scenarios and pipes them into a multiclass classification problem. The BCI architecture is made up by (i) the acquisition unit, (ii) the processing unit and (iii) the navigation unit. The former unit allows collecting EEG data by 6 smart wireless electrodes from the parietal-cortex area from a 32-channel wireless EEG helmet. The processing unit is composed of a dedicated µPC (Raspberry Pi or PC) performing stimuli delivery, data gathering, Machine Learning (ML) and real-time multidimensional classification leading to the user intention recognition. The ML stage is based on a custom algorithm (t-RIDE) which trains the following classification stage on the user-tuned P300 reference features. The ML algorithm starts with a fast calibration phase (just ~190s for the first learning) and comprises a functional approach of feature extraction based on “cognitive chronometry”. The extracted features undergo a dimensionality reduction step and then, are used to define proper decision boundaries for the real-time classification. The real-time classification performs a functional approach for time-domain features extraction, reducing the amount of data to be analysed and speeding up the responses. The Raspberry-based navigation unit actuates the received commands and supports the mechatronic devices by using peripheral sensors. A proof of concept of the proposed BCI has been realized using a prototype car, tested on 7 subjects (aged 26 ± 3). The experimental results show that the novel ML approach allows a complete P300 spatio-temporal characterization in 1.95s using 38 target brain visual stimuli (for each direction of the car path) and then it is ready to drive a mechatronic device. In free-drive mode (testing set), the BCI classification reaches 84.28 ± 0.87% on single-trial detection accuracy. In the worst-case computational time is 19.65ms ± 10.1. The BCI system here described can be also used on different mechatronic actuators, such as robots.
Additional material
- Presentation slides: [pdf]