Yanru Bai

Gait recognition based on multiple views fusion of wavelet descriptor and human skeleton model

Gait recognition is a relatively new subfield in biometric recognition, which attempts to recognize people from the way they walk or run. This paper discusses silhouette-based feature descriptor. Human silhouette geometry is generated by boundary tracking approach and resampled to a normalized format. Boundary-centroid distance is proposed to describe gait modality. Then, we apply wavelet transform to boundary-centroid distance, and extract wavelet descriptor. At the same time, we obtain the human skeleton model and extract bodys dynamic parameters to express gait modality. We carry out human identification based on SVM using the two kinds of gait feature. The performances based on the two features are compared. Multiple feature fusion and multiple views fusion are carried out and the recognition results demonstrate that the performance of multiple features and multiple views recognition is better than any single feature and single view recognition.

A gait stability investigation into FES-assisted paraplegic walking based on the walker tipping index

The gait outcome measures used in clinical trials of paraplegic locomotor training determine the effectiveness of improved walking function assisted by the functional electrical stimulation (FES) system. Focused on kinematic, kinetic or physiological changes of paraplegic patients, traditional methods cannot quantify the walking stability or identify the unstable factors of gait in real time. Up until now, the published studies on dynamic gait stability for the effective use of FES have been limited. In this paper, the walker tipping index (WTI) was used to analyze and process gait stability in FES-assisted paraplegic walking. The main instrument was a specialized walker dynamometer system based on a multi-channel strain-gauge bridge network fixed on the frame of the walker. This system collected force information for the handle reaction vector between the patients upper extremities and the walker during the walking process; the information was then converted into walker tipping index data, which is an evaluation indicator of the patients walking stability. To demonstrate the potential usefulness of WTI in gait analysis, a preliminary clinical trial was conducted with seven paraplegic patients who were undergoing FES-assisted walking training and seven normal control subjects. The gait stability levels were quantified for these patients under different stimulation patterns and controls under normal walking with knee-immobilization through WTI analysis. The results showed that the walking stability in the FES-assisted paraplegic group was worse than that in the control subject group, with the primary concern being in the anterior-posterior plane. This new technique is practical for distinguishing useful gait information from the viewpoint of stability, and may be further applied in FES-assisted paraplegic walking rehabilitation.

Novel gait recognition technique based on SVM fusion of PCA-processed contour projection and skeleton model features

Gait is a potential behavioral feature, and many allied studies have demonstrated that it can be served as a useful biometric feature for recognition. This paper described a novel gait recognition technique based on support vector machine fusion of contour projection and skeleton model features. A principal component analysis method was used to lower the dimension of contour projection after segmenting silhouettes from the background in the key frame of gait picture sequence and a skeleton model was built to produce other shape features. The combining features were fused by a support vector machine and tested on the CASIA database at the feature level and decision level based on posterior probability. Experimental results have demonstrated the effectiveness and advantages of the proposed algorithm.

Individual Feature Extraction and Identification on EEG Signals in Relax and Visual Evoked Tasks

Compared to conventional biometrics, electroencephalogram (EEG) signal has obvious advantages in uniqueness, high confidentiality and impossibility to steal or mimic. In this paper, we investigated EEG signals in relax task and visual evoked task and compared their potentials as the biometric authentication feature. 20 subjects were recruited, and each performed two tasks while 64-channel EEG signals were recorded continuously. The extracted features, autoregression (AR) model, power spectrum of the time-domain (TPS), power spectrum of the frequency-domain (FPS) and phase-locking value (PLV), were given to a support vector machine (SVM) for classification respectively. The results showed that visual evoked task presented better performance in identifying the individuals than the relax task did. Specially, among all these features, AR model got the highest accuracy in both tasks, achieving 90.53% and 96.25% respectively for relax task and visual evoked task. Then support vector machine-recursive feature elimination (SVM-RFE) was employed to select the most discriminative channels just for AR model based on VEP signals for it showed the best performance. Additionally, it gave a higher accuracy of 97.25% based on the 32 top ranked channels. Further investigation may help develop an alternative EEG based biometric system to enhance the traditional biometric technologies.

ICA-SVM combination algorithm for identification of motor imagery potentials

Mental tasks such as motor imagery in synchronization with a cue which result event related desynchronization (ERD) and event related synchronization (ERS) are usually studied in brain-computer interface (BCI) system. In this paper we analyze and classify the ERD/ERS response evoked by the motor imagery of left hand, right hand, foot and tongue. The signals were spatially filtered by Independent Component Analysis (ICA) before calculating the power spectral density (PSD) for related electrodes, and then the Support Vector Machine (SVM) was adopted to recognise the different imagery pattern according to ERD/ERS feature for the signals. The results showed that the combination of ICA-based signal extraction algorithm and SVM-based classification method was an effective tool for the identification of motor imagery potentials, with the highest accuracy rate of 91.4% and 77.6% for the lowest.

Feature Extraction of Individual Differences for Identification Recognition Based on Resting EEG

Biometric recognition based on individual difference was commonly used in many aspects in life. Compared with the traditional features used in person identification, EEG-based biometry is an emerging research topic with high security and uniqueness, and it may open new research applications in the future. However, little work has been done within this area. In this paper, four feature extraction techniques were employed to characterize the resting EEG signals: AR model, time-domain power spectrum, frequency-domain power spectrum and phase locking value. In our experiments using 20 healthy subjects, the classification accuracy by support vector machine reached 90.52% with AR model parameters, highest of the four kinds of features. The results show the potential applications of resting EEG signal in person identification.

Brain-computer interface technique for electro-acupuncture stimulation control

Electro-acupuncture stimulation (EAS) technique applies the electrical nerve stimulation therapy on traditional acupuncture points to restore the muscle tension. The rapid rise and development of brain-computer interface (BCI) technology makes the thought-control of EAS possible. This paper designed a new BCI-controls-EAS (BCICEAS) system by using event related desynchronization (ERD) of EEG signal evoked by imaginary movement. The Fisher parameters were extracted from feature frequency bands of EEG and classified into EAS control commands by Mahalanobis Classifier. A feedback training technique was introduced to enhance the signal feature through a visual feedback interface with a virtual liquid column, which height varied along with EEG power spectral feature. Experimental results demonstrated the validity of the proposed method, including the effective improvement of feedback training on signal feature and reliable control of EAS. It is hoped the BCICEAS can explore a new way for EAS system design and help people who sufferers with severe movement dysfunction.

Super resolution reconstruction based on total variation regularization

Super resolution reconstruction is an important branch of image processing that extracting high resolution images containing more details from an image sequence of low resolution, by image processing such as motion estimation, de-blurring and de-noising. Currently super resolution is an economical and practical algorithm that can be used to improve image resolution in remote monitoring, remote sensing and medical imaging. In this thesis, in order to obtain high resolution image from an image sequence of low resolution and improve the image quality, visual effects, total variation algorithm is used to estimate the motion of low resolution images caused by the restriction of environmental conditions and the physical limitations of imaging equipment. This algorithm contains a lot of processing technologies, such as, motion estimate, motion compensation, image fusion, de-nosing. Experiment result shows that the entropy of the high resolution image was improved and the D and Dindex are improved with the increasing of frames, so clearly high resolution image can be obtained from source image by using this algorithm. The super resolution algorithm mentioned in the thesis with high practical application value, can be applied to long-range remote sensing and face image restoration.