Nikolay Neshov

Pain detection from facial characteristics using supervised descent method

Current methods of assessing pain depend almost entirely on verbal report such as clinical interview or questionnaires of the patients. Pain being a symptom that can neither be felt nor seen, poses a major problem for the medical personnel involved in pain management since there are no accurate objective measures to establish the extent of pain the patient is suffering from when using a remote assistive medical system. The verbal report grading pain has obvious discrepancies, especially when it comes to children or people with limited ability to communicate (i.e. the mute, mentally impaired, or patients having assisted breathing). When designing a medical assistive system measuring pain in an efficient way is of great importance. In this paper we proposes an algorithm for both automatic pain recognition (i.e. pain/no pain presence in human) and continuous pain intensity estimation based on facial expression analysis. To locate specific landmarks in the face we used Supervised Descent Method (SDM) and then extract feature vectors using Scale Invariant Feature Transform (SIFT). For the recognition task we build a classier based on Support Vector Machines (SVM) and for the continuous pain intensity estimation task we trained linear regressor. The experiments with patients with shoulder pain show very good recognition rate (more than 95.7%). For the pain intensity estimation we reached an average Mean Squared Error of 1.28 and Correlation coefficient of 0.59. The recorded results demonstrate performance that exceeds state-of-the-art results on a standard data set.

Comparison on Late Fusion Methods of Low Level Features for Content Based Image Retrieval

Finding the right feature for image representation is an important key to attaining successful Content Based Image Retrieval (CBIR) system. This choice depends on the content of images to be searched. Todays real world image databases are heterogeneous and consist of images that can be described appropriately using different feature types. One approach to deal with this is to utilize late fusion methods. That is, the CBIR system must be able to fuse multiple results produced by each feature. In this paper by experimental comparison of output results achieved from eleven low-level features over three image databases an appropriate several sets of features are selected and five late fusion methods are applied over each set. By analysis of the results for all methods it has been shown which ones reach the best performances and stable retrieval accuracy among the investigated image databases and sets of features.

Facial Expression Classification Using Supervised Descent Method Combined With PCA and SVM

It has been well known that there is a correlation between facial expression and person’s internal emotional state. In this paper we use an approach to distinguish between neutral and some other expression: based on the displacement of important facial points (coordinates of edges of the mouth, eyes, eyebrows, etc.). Further the feature vectors are formed by concatenating the landmarks data from Supervised Descent Method, applying PCA and use these data as an input to Support Vector Machine (SVM) classifier. The experimental results show improvement of the recognition rate in comparison to some state-of-the-art facial expression recognition techniques.

Classification of Mental Tasks from EEG Signals Using Spectral Analysis, PCA and SVM

Abstract Signals provided by the ElectroEncephaloGraphy (EEG) are widely used in Brain-Computer Interface (BCI) applications. They can be further analyzed and used for thinking activity recognition. In this paper we proposed an algorithm that is able to recognize five mental tasks using 6 channel EEG data. The main idea is to separate the raw EEG signals into several frames and compute their spectrums. Next, a second-order derivative of Gaussian is applied to extract features and an optimum Gaussian kernel parameters grid search is performed with the help of cross-validation. The extracted features are further reduced by Principal Component Analysis. The processed data is utilized to train SVM classifier which is used for mental tasks recognition afterwards. The performance of the algorithm is estimated on publically available dataset. In terms of 5 folds cross-validation we obtained an average of 82.7% recognition rate (accuracy). Additional experiments were conducted using leave-one-out cross-validation where 67.2% correct classification was reported. Comparison to several state-of-the art methods reveals the advantages of the proposed algorithm.

Drowsiness monitoring in real-time based on supervised descent method

With increased work load and unsuitable work shifts to survive in the fast paced world of today, people tend to lose sleep. Irregular sleep patterns and lack of sleep leads to drowsiness and fatigue. Drowsiness is perilous for the driver himself and for other drivers on the road and must be avoided for example by noise alerts in the car. This paper describes a method to detect drowsiness after implementing eye-tracking and mouth shape tracking in real-time. Viola-Jones algorithm is used to detect facial features in real-time. The proposed approach uses the detected facial features (i.e. eyes and mouth) based on Supervised Descent Method to find the blinking rate of a driver as well as for yawning detection. A decision, whether the driver is vigilant or not is then provided. Real time experiments based on publicly available dataset prove that the proposed method is highly efficient in finding the drowsiness and alerting the driver.

Combined EEG and EMG fatigue measurement framework with application to hybrid brain-computer interface

In recent years, the EEG-based brain-computer interface (BCI) has become one of the most promising areas of research in computer science and robotics. Many internationally renewned research teams combining engineers and doctors, experts in neuroscience are trying to develop useful applications and devices offering disabled people to lead a normal life. Useful BCIs for disabled people suffering from Cerebral palsy, Parkinsons disease, Brain injury, Spinal cord injuries, Multiple sclerosis, Stroke, Post-polio syndrome should allow them to use all their existing brain and muscle abilities as control possibilities. In this paper we present a framework based on the mutimodal fusion approach of the users electromyographic (EMG) and electroencephalographic (EEG) activities in a so called “Hybrid-BCI” (hBCI). Although EEG BCI alone yields good performance as already proved in many research papers, it is outperformed by the fusion of EEG and EMG. We investigate the influence of muscular fatigue on the EMG performance. Such a framework will allow a more reliable control and adaptation of the hBCI if the user get exhausted and loses concentration during the rehabilitation process. We focus our research in aims of improving the lives of many upper limb disabled individuals through a combination of current BCI technologies with existing assistive medical systems.

Automatic Pain Detection in Video Sequences for Neuro-Rehabilitation

Adaptive and interactive mental engagement combined with positive emotional state are requirements for an optimal outcome of the neuro-rehabilitation process for patients with brain damage usually caused by TBI (traumatic brain injury), stroke or brain disease such as cancer, epilepsy, and Alzheimers disease. We propose a method for automatic pain recognition in video sequences using the landmarks data from Supervised Descent Method and applying Support Vector Machine (SVM) for data classification. This method is suitable for being part of assistive medical system for neuro-rehabilitation of patients with TBI. The experiments with a video dataset with patients with shoulder pain show very good recognition rate (95,7%) for recognizing the painful facial states of the subjects.

Finding URLs in images by text extraction in DCT domain, recognition and matching in dictionary

Automatic and fast detection of overlaid URLs and text in images and video with adult content can be an important aid for children protection in the internet. In this paper, we propose a method and ...

Combining Features Evaluation Approach in Content-Based Image Search for Medical Applications

In this paper we propose an approach for a feature combination helping to distinguish searched images from databases by retrieving relevant images. The retrieval effectiveness of 11 well known image features, commonly used in Content Based Image Retrieval (CBIR) systems, is investigated. We suggest a combined features approach including features’ performance comparison of 57 various medical image categories from IRMA Database. The most informative 3 features, adaptive to image categories, are defined. Based on experiments and image similarity accuracy analysis we suggest a set of 3 low level features Color Layout, Edge Histogram and DCT Coefficients. The developed approach achieves better similar images retrieval results for more image classes. The results show an accuracy improvement of 14.49% on Mean Average Precision (MAP). The comparison is done to the same type performance measure of the best individual feature in different medical image categories.