Siamak Rezaei

Different classification techniques considering brain computer interface applications.

In this work the application of different machine learning techniques for classification of mental tasks from electroencephalograph (EEG) signals is investigated. The main application for this research is the improvement of brain computer interface (BCI) systems. For this purpose, Bayesian graphical network, neural network, Bayesian quadratic, Fisher linear and hidden Markov model classifiers are applied to two known EEG datasets in the BCI field. The Bayesian network classifier is used for the first time in this work for classification of EEG signals. The Bayesian network appeared to have a significant accuracy and more consistent classification compared to the other four methods. In addition to classical correct classification accuracy criteria, the mutual information is also used to compare the classification results with other BCI groups.

Integrating wireless EEGs into medical sensor networks

Wireless Electroencephalograms (EEG) are currently being used to wirelessly transmit the data from brain sensors to a computer and they carry huge potential for many future medical applications. This paper presents the design of a hybrid medical sensor network with Tmote Sky motes as wireless EEG sensor nodes at the lowest level collecting EEG signals and sending them to Stargate PDAs at the next level. Stargates perform artifact removal, Fourier transformation and feature extraction and the final machine intelligence algorithms are run at a PC server. Several features of the CodeBlue medical sensor network like query processing, routing layer are used in our design. The advantages of the hybrid medical sensor network integrating wireless EEGs include the capability to have the brain monitoring functionality incorporated into the medical sensor networks.

Selecting better EEG channels for classification of mental tasks

In this work a new method is proposed to reduce the number of EEG channels needed to classify mental tasks. By applying genetic algorithm to the search space consisting of 6 channel combinations of 19 EEG channels the more salient combinations of them in classification of three mental tasks are selected. This algorithm reduces the calculation time and the final results are verified by our observations. Obtained results bring forward the concept of systematic and intelligent selection criteria for choosing superior EEG channels of subjects for mental task classification. This may find applications in the field of brain computer interfaces which are based on classifications of mental tasks, by reducing the number of EEG channels.

Appearance-based Pain Recognition from Video Sequences

In this paper, we present an appearance-based approach for pain recognition from video sequences. An automatic face detector is employed which uses skin color modeling to detect human face in the video sequence. The pain affected portions of the face are obtained by using a mask image. The obtained face images are then projected onto a feature space, defined by eigen-faces, to produce the biometric template. Recognition is performed by projecting a new image onto the feature spaces spanned by the eigen-faces and then classifying the painful face by comparing its position in the feature spaces with the positions of known individuals. To check better accuracy, the system is tested against two more feature spaces defined by eigen-eyes and eigen-lips.

Comparison of Five Different Classifiers for Classification of Mental Tasks

In this work we consider classification of mental tasks from EEG signal by using five different classifiers. These include neural network, Bayesian graphical network, Bayesian quadratic classifier, hidden Markov model and Fisher linear classifier. The experimental results are compared to each other. The Bayesian network appears to have comparable accuracy with the best classifier of the five but the classification time for it is more than others

A Real-Time Face Recognition Approach from Video Sequence using Skin Color Model and Eigenface Method

We present a robust approach for real time face recognition from video sequences. An automatic face detector is employed which uses skin color modeling to detect human face in the video sequence. The presence or absence of face in each region is verified by means of height-width proportion and an eye detector: based on efficient template matching scheme. The obtained face images are then projected onto a feature space, defined by eigenfaces, to produce the biometric template. Recognition is performed by projecting a new image onto the feature spaces spanned by the eigenfaces and then classifying the face by comparing its position in the feature spaces with the positions of known individuals. The proposed system can be used in security purposes and in any visual communication system, such as teleconferencing, human computer interaction etc

Video analysis for view-based painful expression recognition

In recent years, facial expressions of pain have been the focus of considerable behavioral research. Such work has documented that pain expressions, like other affective facial expressions, play an important role in social communication. Enabling computer systems to recognize pain expression from facial images is a challenging research topic. In this paper, we present two systems for pain recognition from video sequences. The first approach, eigenimage, projects the face images, detected from video sequences, onto a feature space, defined by eigenfaces, to produce the biometric template. Recognition is performed by projecting a new image onto that feature space and then classifying the face by comparing its position in the feature spaces with the positions of known individuals. To ensure better accuracy, the system is tested against two more feature spaces defined by eigeneyes and eigenlips. The second approach, neural network, extracts location and shape features of the detected faces and uses them as inputs to the artificial neural network which employs the standard error back-propagation algorithm for classification of faces. From experiments, we conclude that neural network based method is better in terms of speed and accuracy than eigenimage based method.

Divide-and-Conquer Algorithm for Clustalw-MPI

Multiple sequence alignment continues to be an active field of research in computational biology and the most widely used tool for multiple sequence alignment is ClustalW, which achieves alignment via three steps: pair wise alignment, guide tree generation and progressive alignment. ClustalW-MPI is a parallel implementation of ClustalW. In this paper, a new approach, divide-and-conquer, is implemented which uses ClustalW-MPI for sequence alignment but it gets a better speed up performance than ClustalW-MPI. In this approach, the sequences are first cut down into smaller subsequences by divide-and-conquer technique to minimize the computational space. Then these subsequences are sent to different available processors using message passing interface technique. Those processors align the subsequences by executing ClustalW-MPI simultaneously. After aligning, the results are then sent to the main processor to be concatenated to produce the final alignment. But some quality of the alignment may be compromised in this approach for the introduction of gaps at the start or end of subsequences aligned. Therefore, some heuristic methods for fixing the cut points were suggested for future improvement, such as overlapping alignment and sliding window alignment

Support vector machine for automatic pain recognition

Facial expressions are a key index of emotion and the interpretation of such expressions of emotion is critical to everyday social functioning. In this paper, we present an efficient video analysis technique for recognition of a specific expression, pain, from human faces. We employ an automatic face detector which detects face from the stored video frame using skin color modeling technique. For pain recognition, location and shape features of the detected faces are computed. These features are then used as inputs to a support vector machine (SVM) for classification. We compare the results with neural network based and eigenimage based automatic pain recognition systems. The experiment results indicate that using support vector machine as classifier can certainly improve the performance of automatic pain recognition system.

Performance Comparison of MPI-Based Parallel Multiple Sequence Alignment Algorithm Using Single and Multiple Guide Trees

Multiple alignment of biological sequences is an interesting area of research for application of parallel processing algorithms. For multiple alignment of sequences, parallelism can be introduced at different levels to reduce the overall time. This paper is concerned with the comparison of an MPI-based multiple sequence alignment algorithm by using single and multiple guide trees. In this work, we will look at the application of divide-and-conquer algorithm for multiple sequence alignment and discuss two different implementations on MPI. We compare the speed up of the two approaches and we contrast the quality of the alignments for these two approaches. These two approaches differ in using one single tree for alignment vs. multiple trees for alignment that we further discuss in this paper