Drowsiness is a condition between waking and sleeping. Drowsiness is also one of the factors causing road accidents. Data on traffic accidents shows that 83% of events are caused by fatigue, drowsiness, and driving the speed threshold.
Therefore, an alarm system is needed to prevent drowsiness while driving. Previously, many alarm systems have been developed that use cameras to monitor eye movements or sensors placed on tires to prevent slippage when the driver is drowsy. But this system is still limited due to the intensity of light or weather. Another alternative offered to overcome this problem is to use brain waves recorded by EEG (Electroencephalograph) based on BCI (Brain Computer Interface).
The brain is the most interesting organ to study or observe because all body commands are coordinated in the brain. Brain signals are complex signals that can be produced by the human body. It consists of electrode channels which each represent electrical activity from parts or regions in the brain.
BCI is not like an EEG device in a hospital with many cables connected but rather shaped like a wireless headset connected via Bluetooth so it is easy to use. BCI uses the EEG principle and has been developed commercially with affordable prices for games and medical applications. One type of commercial BCI is EMOTIV EPOC + by having 14 electrode channels: AF3, AF4, F3, F4, F7 and F8 representing frontal cortex, T7-T8 represent the right and left temporal cortex, P7-P8 represents the parietal cortex and O1-O2 represents occipital cortex. BCI can also be used to record electrical activities of the brain when drowsy and awake condition.
When there is drowsiness, alpha, and theta waves, intensity increases compared to when the condition is awake. Alpha waves are brain waves with a frequency of 8-12 Hz, while theta waves have a frequency of 4-8 Hz. Both appear in normal people when they are relaxed and tend to be drowsy. It can be used to detect drowsiness when driving. With the help of wavelets as a processing method to convert signals of time into frequency forms. It is supported with Artificial Neural Network method, Extreme Learning Machine to distinguish drowsiness and wakefulness automatically.
Extreme Learning Machine is a popular method that can classify groups based on the characteristic features of each group quickly and with high accuracy. The extreme learning machine has two stages, including the training process and the testing process. The training process is important in the Extreme learning machine because the trained features will determine the test results at the end. If the feature entered does not represent the characteristics of each class, the test results will have a low accuracy value. Although there is a difference in frequency between drowsiness and wakefulness, the increasing in alpha and theta waves during drowsiness, the difference is not statistically significant so that it will reduce the accuracy of Extreme Learning Machine testing. Therefore, it is necessary to add the Common Spatial Pattern method, which can increase the difference in values of the two groups.
Analysis and classification system based on extreme learning machine by utilizing Delta brain wave frequency features (0-4 Hz), Theta (4-8 Hz), Alfa (8-16 Hz) and Beta (16-32 Hz) which are further processed using the Common Spatial Pattern method obtains accuracy above 91%. Whereas the classification results of the Extreme learning machine by utilizing the brain wave frequency feature without adding the Common Spatial Pattern method only gained an accuracy of 72%. It shows that the Extreme learning machine supported by the Common Spatial Pattern can be used to distinguish EEGs in drowsiness and awake conditions. This system is expected to be useful in automatic alarm system development with a brainwave signal base to detect sleepy motorists. So it can reduce the risk of accidents.
Author : Osmalina Nur Rahma
Details of the research available at:
http://jmss.mui.ac.ir/index.php/jmss/article/view/489
Osmalina Nur Rahma, Akif Rahmatillah. Drowsiness Analysis Using Common Spatial Pattern and Extreme Learning Machine Based on electroencephalogram Signal. Journal of Medical Signals and Sensors, ISSN : 2228-7477
