BCI Applications For Drivers | Drowsiness Detection Using BCI and EEG

 

8 Methods of Drowsiness Detection in drivers using BCI and EEG

 

Tensor-Based EEG Network Formation and Feature Extraction for Cross-Session Driving Sleepiness Detection

EEG feature extraction method called Tensor Network Features TNF is used to manipulate  structure of Sleepiness patterns and extract features based on tensor network. 

Some priviously proposed methods for Driving Sleepiness detection are spectral power features and fuzzy entropy features. They consider the features in each channel separately to identify Sleepiness so they lack sufficient data.

 This TNF method introduces Tucker decomposition to tensorized EEG channel data of training set, then features of training and testing tensor samples are extracted from the corresponding subspace matrices through tensor network summation.

 Compared with spectral power features and fuzzy entropy features, the accuracy of TNF method is improved by 6.7% and 10.3% on average with maximum value 17.3% and 29.7% respectively, which is promising in developing practical and robust cross-session driving Sleepiness detection system.

FUZZY DIVERGENCE BASED ANALYSIS FOR EEG Sleepiness DETECTION BRAIN-COMPUTER INTERFACES

Low signal to noise ratio and non-stationarities in EEG signals is one of the leading challenges in Brain-Computer interface. We deal with this by using machine learning regression to develop novel spatial filtering based feature extraction methods to detect Sleepiness . 

This method is called DivOVR-FuzzyCSP-WS based features which outperformed fuzzy CSP based baseline features in terms of both RMSE and CC performance metrics. 

This feature extraction method will help researchers in developing algorithms for signal processing and Brain-Computer interface regression problems.

Time Domain Parameters as a feature for single-channel EEG-based Sleepiness detection method

Several feature extraction methods have been used in EEG-based Sleepiness detection. 

Here is a novel feature extraction strategy based on a single Hjorth parameter, and compared its classification capability with the existing Power spectral density (PSD) feature extarction method.  

The results show that the proposed H-parameter features have higher and stronger performance compared to the PSD. This is the first time we practically apply Hjorth parameters to EEG-based driver Sleepiness detection.

Driver drowsiness Detection with Single EEG Channel Using Transfer Learning

EEG signal is one of the reliable physiological signals used to perceive driver Sleepiness state but wearing a multi-channel headset is difficult for drivers. We can use a driver drowsiness detection system using transfer learning, depending only on one EEG channel to increase system usability. 

The system firstly acquires the signal and passing it through preprocessing filtering then, converts it to a 2D spectrogram. Finally, the 2D spectrogram is classified with AlexNet using transfer learning to classify it either normal or drowsiness state.

We compared the accuracy of seven EEG channel to select one of them as the most accurate channel to depend on it for classification.  

The results show that the channels FP1 and T3 are the most effective channels to indicate the drive drowsiness state. They achieved an accuracy of 90% and 91% respectively. This will result in an efficient driver drowsiness detection system.

EEG signal cleaning for Sleepiness detection

At least one EEG signal recorder and an algorithm must be developed to interprete the brain activity. In this study two new adapted algorithms are proposed for EEG signal cleaning and also some basic features will be extracted in order to detect Sleepiness. 

We used physionet database in order to perform the task.  The algorithms adapt zero-crossing filter for power supply noise removal and also the usage of the polynomial interpolation to remove the baseline artifact.

Multimodal System to Detect Driver drowsiness Using EEG, Gyroscope, and Image Processing

There are many Sleepiness detection methods like EEG based, image acquisition technology and bio-mathematical models. However there has been less research on hybrid of these systems.

There is another Neural network based hybrid multi modal system that detects driver drowsiness using electroencephalography EEG data, gyroscope data and image processing data. 

It was found that this hybrid system performed well with a detection accuracy of 93.91% in identifying the Sleepiness state of the driver.

Electroencephalogram Based Reaction Time Prediction With Differential Phase Synchrony Representations Using Co-Operative Multi-Task Deep Neural Networks

EEG-based reaction time prediction and Sleepiness detection are considered as multi-task learning problems. Here is another driving Sleepiness detection method which utilizes the fuzzy common spatial pattern optimized differential phase synchrony representations to inspect EEG synchronization changes from the alert state to the drowsy state.

 This method can be used to distinguish between alert and drowsy state of mind using statistical analysis. The proposed Multi-Task DeepNet MTDNN performs superior to the baseline regression schemes, like:

• Support vector regression SVR
• Least absolute shrinkage and selection operato
• Ridge regression
• K-nearest neighbors
• Adaptive neuro fuzzy inference scheme ANFIS

In terms of root mean squared error RMSE, mean absolute percentage error MAPE, and correlation coefficient metrics. 

The best performing multi-task network MTDNN5 recorded a 15.49% smaller RMSE, a 27.15% smaller MAPE, and a 10.13% larger CC value than SVR.

An Effective Hybrid Model for EEG-Based Sleepiness Detection

In this study we proposed a hybrid EEG-based Sleepiness detection system composed of three main building blocks. Both raw EEG signals and their corresponding spectrograms are used in the proposed building blocks.

 In the first building block, while energy distribution and zero-crossing distribution features are calculated from the raw EEG signals, spectral entropy and instantaneous frequency features are extracted from the EEG spectrogram images. 

In the second building block, deep feature extraction is employed directly on the EEG spectrogram images using pre-trained AlexNet and VGGNet.

 In the third building block, the tunable Q-factor wavelet transform TQWT is used to decompose the EEG signals into related sub-bands. The spectrogram images of the obtained sub-bands and statistical features, such as mean and standard deviation of the sub-bands’ instantaneous frequencies, are then calculated. 

Each feature group from each building block is fed to a long-short term memory LSTM network for the purposes of classification. The obtained results from the LSTM networks are then fused with a majority voting layer. 

The MIT-BIH Polysomnographic database was used in the experimental works. The evaluation of the proposed method was carried out with ten-fold cross validation test and the obtained average accuracy score was 94.31%. The method’s achievement was found to be better than the compared results.

For more interesting research regarding Brain-Computer interface and BCI projects please visit BCIProjects blog.