Kernel Support Vector Data Description Research Articles

Face Recognition Based on Videos by Using Convex Hulls

A wide range of face appearance variations can be modeled by using set-based recognition approaches effectively, but computational complexity of current methods is highly dependent on the set and class sizes. This paper introduces new video-based classification methods designed for reducing the required disk space of data samples and speed up the testing process in large-scale face recognition systems. In the proposed method, image sets collected from videos are approximated with kernelized convex hulls and it was shown that it is sufficient to use only the samples that participate in shaping the image set boundaries in this setting. The kernelized support vector data description (SVDD) is used to extract those important samples that form the image set boundaries. Moreover, we show that these kernelized hypersphere models can also be used to approximate image sets for classification purposes. Then, we propose a binary hierarchical decision tree approach to improve the speed of the classification system even more. At last, we introduce a new video database that includes 285 people with 8 videos of each person, since the most popular video data sets used for set-based recognition methods include either a few people, or small number of videos per person. The experimental results on varying sized databases show that the proposed methods greatly improve the testing times of the classification system (we obtained speed-ups to a factor of 20) without a significant drop in accuracies.

Automated Detection of Convulsive Seizures Using a Wearable Accelerometer Device.

Epileptic seizure detection requires specialized approaches such as video/electroencephalography monitoring. However, these approaches are restricted mainly to hospital setting and requires video/EEG analysis by experts, which makes these approaches resource- and labor-intensive. In contrast, we aim to develop a wireless remote monitoring system based on a single wrist-worn accelerometer device, which is sensitive to multiple types of convulsive seizures and is capable of detecting seizures with short duration. Simple time domain features including a new set of Poincar´e plot based features were extracted from the active movement events recorded using a wrist-worn accelerometer device. The best features were then selected using the area under the ROC curve analysis. Kernelized support vector data description (SVDD) was then used to classify non-seizure and seizure events. The proposed algorithm was evaluated on 5;576h of recordings from 79 patients and detected 40 (86:95%) of 46 convulsive seizures (generalized tonic-clonic (GTCS), psychogenic non-epileptic (PNES), and complex partial seizures (CPS)) from twenty patients with a total of 270 false alarms (1:16=24h). Furthermore, the algorithm showed a comparable performance (sensitivity 95:23% and false alarm rate 0:64=24h) with respect to existing unimodal and multi-modal methods for GTCS detection. The promising results shows the potential to build an ambulatory monitoring convulsive seizure detection system. A wearable accelerometer based seizure detection system would aid in continuous assessment of convulsive seizures in a timely and non-invasive manner.

A novel multimode process monitoring method integrating LDRSKM with Bayesian inference

A local discriminant regularized soft k-means (LDRSKM) method with Bayesian inference is proposed for multimode process monitoring. LDRSKM extends the regularized soft k-means algorithm by exploiting the local and non-local geometric information of the data and generalized linear discriminant analysis to provide a better and more meaningful data partition. LDRSKM can perform clustering and subspace selection simultaneously, enhancing the separability of data residing in different clusters. With the data partition obtained, kernel support vector data description (KSVDD) is used to establish the monitoring statistics and control limits. Two Bayesian inference based global fault detection indicators are then developed using the local monitoring results associated with principal and residual subspaces. Based on clustering analysis, Bayesian inference and manifold learning methods, the within and cross-mode correlations, and local geometric information can be exploited to enhance monitoring performances for nonlinear and non-Gaussian processes. The effectiveness and efficiency of the proposed method are evaluated using the Tennessee Eastman benchmark process.