g.PHYSIOobserver: Specs & Features
Analyze emotions, workload, activity based on physiology
The g.PHYSIOobserver is a complete system to classify different states of a subject based on physiological parameters. The system contains all necessary components to quantify emotions, workload, physical tasks and many other things. The g.PHYSIOobserver works with many different sensors and electrodes to measure physiological and physical parameters of a subject and can calculate many different parameters from these signals. A key feature of the system is that it allows you to run experimental paradigms that are synchronized with physiological signals. The paradigms allow you to bring the subject into specific states of emotions, workload, memory tasks, etc., while all parameters are captured. These states can be chosen by the experimenter. Then, a classification algorithm is trained on these parameters during the different states and tries to discriminate them. Finally, the accuracy is calculated and gives an objective measure of the quality of the classification. The g.PHYSIOobserver works also in real-time and can therefore track the current state of a subject on-line. This information can be transmitted to other applications or devices, including real-time feedback systems.
- Train the g.PHYSIOobserver with different tasks
- Classifies physiological parameters in real-time to determine the state of the subject
- Gives accuracy as an objective measure
- Select from a large variety of different parameters
- Send the classification result to other applications to execute closed-loop experiments
The g.PHYSIOobserver is able to measure ECG, EEG, EMG, GSR, respiration, temperature, acceleration and oxygen saturation with g.USBamp or g.MOBIlab+. This biosignal data is transmitted via USB or wireless to the recording computer that is storing and visualizing the data for inspection. The recording computer also controls the experimental paradigm that instructs the subject about different tasks (e.g. calculating). The real-time processing system extracts parameters from the biosignal data such as heart rate, heart rate variability, respiration rate, inhalation time, change rate of GSR, etc. and classifies the data. Finally, the classification result predicts the subject's current state. This result is updated in real-time, and can also be transmitted to other applications or the experimenter.
The experimental paradigms are presented by default on a computer screen that gives the instructions to the subject. You can also use a head-mounted device, the g.VRsys, a Virtual Reality system from g.tec, or a custom exoskeleton system. Furthermore, the system can work with eye- and movement-trackers, to give tone, electrical or tactile stimulation. A microphone can be connected to log subject responses.
The UDP interface allows you to send the classification result, but also the calculated parameters to other applications to build up real-time loops.
The first task is to place all the electrodes and sensors on the subject. Then, the experimental paradigm begins, during which the subject receives instructions from the computer screen and has to perform specific predefined tasks. Meanwhile, all the subject's biosignal data is acquired along with synchronization triggers. When the paradigm stops, the off-line analysis begins to calculate parameters. With these parameters, a classification algorithm is trained which discriminates between the different states of the experimental paradigm and gives an accuracy level to quantify the separability. If the accuracy is not good enough then (i) additional sensors or electrodes can be used, (ii) the paradigm can be improved, or (iii) additional parameters can be added. One good parameter for detecting mental counting is the bandpower of EEG data in the alpha region. When the accuracy meets the expectations of the experimenter, then the g.PHYSIOobserver is ready for real-time tests. The paradigms are started again, and now the system gives a real-time prediction of the current state of the subject with a certain likelihood. The parameters, as well as the predicted state, can be sent to other applications. Finally, the accuracy can be calculated again to see if the system can discriminate the different states.