Increase Recognition Accuracy Research Articles

基于粒子群优化的条件概率神经网络

条件概率神经网络在进行模式分类时具有独特的优势，然而如何对其进行有效的训练，从而找到最优参数却是一个困难的问题。在考虑条件概率神经网络的结构特点之后，本文提出了一种基于粒子群优化的条件概率神经网络的训练方法。我们将这种基于粒子群优化的条件概率神经网络用于人脸年龄估计，实验结果表明这种网络能够显著地提高识别的准确率。 Conditional probability neural network (CPNN) has special advantage in pattern classification problems. However, how to find the optimal parameters of the CPNN to achieve better perfor-mance is an extraordinary challenge. Considering the structure feature of CPNN, we proposed a new training method based on particle swarm optimization (PSO). This method utilizes PSO to optimize the structure of CPNN and label distributions by introducing Hellinger distance between different label distributions. We applied the improved CPNN on facial age estimation. The experimental results showed that this network could increase recognition accuracy significantly.

Nonlinear Speech Analysis and Modeling for Malayalam Vowel Recognition

This paper presents a novel method for the recognition of Malayalam vowel phonemes using nonlinear speech parameters such as maximal Lyapunov exponent and Phase Space Anti-diagonal Point Distribution. Reconstructed phase space is used as a base for extracting these features. The results show that the proposed nonlinear features have significant discriminatory power. The combined feature vectors of nonlinear feature and MFCC features yield increased recognition accuracy for five Malayalam vowel phonemes.

FACE RECOGNITION USING LOCAL QUANTIZED PATTERNS AND GABOR FILTERS

Abstract. The problem of face recognition in a natural or artificial environment has received a great deal of researchers’ attention over the last few years. A lot of methods for accurate face recognition have been proposed. Nevertheless, these methods often fail to accurately recognize the person in difficult scenarios, e.g. low resolution, low contrast, pose variations, etc. We therefore propose an approach for accurate and robust face recognition by using local quantized patterns and Gabor filters. The estimation of the eye centers is used as a preprocessing stage. The evaluation of our algorithm on different samples from a standardized FERET database shows that our method is invariant to the general variations of lighting, expression, occlusion and aging. The proposed approach allows about 20% correct recognition accuracy increase compared with the known face recognition algorithms from the OpenCV library. The additional use of Gabor filters can significantly improve the robustness to changes in lighting conditions.

Using web-based training to enhance perceptual-cognitive skills in complex dynamic offside events

In association football, the difficulty in making offside decisions depends on both perceptual and cognitive processes. Therefore, the objectives of the present study were to enhance the decision-making skills of assistant referees by further developing their ability to (1) time slice the incoming information flow into series of isolated time frames during an ongoing offside situation and (2) use this information to mentally read off the spatial positions of the key-role players. Training (n = 10) and control groups (n = 10) were exposed to a pre- and post-test, consisting of an offside decision-making and frame recognition test. In the latter, assistant referees were asked to indicate which of five photos best represented the spatial position of the defender and attacker at the moment of the pass. Only the training group received 12 web-based offside training sessions. First, the training group improved in mentally freezing, holding and scanning the mental picture of the offside situation in short-term memory from pre- to post-test, as evidenced by an increased recognition accuracy. Second, the improvement in recognition accuracy resulted in enhanced performance on the offside decision-making task. The benefits of web-based training are highlighted.

Gait-based person recognition using arbitrary view transformation model.

Gait recognition is a useful biometric trait for person authentication because it is usable even with low image resolution. One challenge is robustness to a view change (cross-view matching); view transformation models (VTMs) have been proposed to solve this. The VTMs work well if the target views are the same as their discrete training views. However, the gait traits are observed from an arbitrary view in a real situation. Thus, the target views may not coincide with discrete training views, resulting in recognition accuracy degradation. We propose an arbitrary VTM (AVTM) that accurately matches a pair of gait traits from an arbitrary view. To realize an AVTM, we first construct 3D gait volume sequences of training subjects, disjoint from the test subjects in the target scene. We then generate 2D gait silhouette sequences of the training subjects by projecting the 3D gait volume sequences onto the same views as the target views, and train the AVTM with gait features extracted from the 2D sequences. In addition, we extend our AVTM by incorporating a part-dependent view selection scheme (AVTM_PdVS), which divides the gait feature into several parts, and sets part-dependent destination views for transformation. Because appropriate destination views may differ for different body parts, the part-dependent destination view selection can suppress transformation errors, leading to increased recognition accuracy. Experiments using data sets collected in different settings show that the AVTM improves the accuracy of cross-view matching and that the AVTM_PdVS further improves the accuracy in many cases, in particular, verification scenarios.

Birds of a feather get misidentified together: High entitativity decreases recognition accuracy for groups of other-race faces

Purpose The cross-race effect can be exaggerated when faces are presented in groups, leading to less accurate eyewitness identifications (Pezdek, O'Brien, & Wasson, 2012, Law Hum. Behav., 36, 488). Our current study examined the effect of entitativity, the degree to which members of a group are perceived as a coherent unit (Campbell, 1958, Behav. Sci., 3, 14), on recognition accuracy for same- and cross-race faces presented in groups. Methods White participants viewed 16 slides of 3-face groups (eight White groups, eight Black groups). Prior to viewing the faces they were told that the entitativity of each 3-face group was high (‘friends who do things together’) or low (‘people in line at the bank’). They were then tested on 32 individually presented faces (16 old and 16 new). Results When cross-race faces were presented in high rather than low entitativity groups, less accurate face recognition memory resulted. Increasing group entitativity decreased recognition accuracy for cross-race faces but increased recognition accuracy for same-race faces. Conclusions The results suggest that the perception of a group negatively impacts eyewitness memory. Contextual factors such as entitativity need to be considered along with other estimator variables when assessing eyewitness identification accuracy.

Non-Invasive Recognition of Poorly Resolved Integrated Circuit Elements

We present a non-invasive method for recognition of components in a digital CMOS integrated circuit (IC). We use a confocal infrared laser scanning optical microscope to collect multimodal images through the backside of the IC. Individual modes correspond to passive reflectivity measurements or active measurements, such as light-induced voltage alteration. The modes are registered and stored in a multidimensional data cube. We apply a machine learning algorithm using a binary representation to identify a variety of data structures from transistors to entire logic cells. Because of the compact representation, objects can be detected rapidly. We show that by increasing the number of imaging modes used to develop the descriptor, we can significantly increase recognition accuracy. The approach allows recognition of poorly resolved components, whose primary distinguishing features are below traditional optical resolution limits, and is general enough to be applied to multiple design processes. We believe this represents a significant step toward a fully non-invasive IC reverse engineering system.

Efficient person authentication based on multi‐level fusion of ear scores

A two-stage geometric approach that is both scale and rotation invariant is implemented for extracting the unique features present in the surface of an ear image. As occlusion because of ear rings and hair significantly affect the efficiency of ear recognition process, only the middle portion of the ear is considered in this work. The resultant matching scores are compared against a threshold to make a decision for authenticating a person. It is found that the fused scores obtained from the two levels of feature extraction enhance the recognition accuracy compared with that of the individual stages. Finally, particle swarm optimisation technique is applied on the matching scores in order to optimise the fusion parameters such as decision threshold and weights. It results in further improved verification rates compared with the fusion of scores without optimisation. Thus, the proposed method works on partial ear images and demonstrate the presence of more unique features in the middle part of the ear (as seen by the increase in recognition accuracy) and the method also aids in reducing the computation time.

Age-related changes in prefrontal and hippocampal contributions to relational encoding

Age-related declines in relational encoding are well documented. It remains unclear, however, whether such declines reflect dysfunction of (1) ventrolateral prefrontal cortex (VLPFC) and deficient generation of associations; and/or (2) hippocampal dysfunction and impoverished binding of associations. In order to separate VLPFC and hippocampal contributions to relational encoding, we manipulated the generative demands of the encoding task by varying the number of semantic associations between the to-be-encoded information (three words). Thus, trials with fewer semantic associations (lower-association trials) require more generative processing during encoding, relative to trials in which more semantic associations are provided for binding (higher-association trials). Parametric modulation analyses on successfully encoded items revealed that, unlike younger adults, older adults did not show an up-regulation of VLPFC activity during lower-association trials. In contrast, hippocampal activity in both older and younger adults was greater in higher- relative to lower-association trials. Moreover, recognition accuracy improved significantly in both groups with the provision of more semantic associations, indicating that both younger and older adults benefitted from this form of encoding support. Our findings suggest that left VLPFC dysfunction may underlie relational encoding deficits in older adults, but that when provided with associations to bind, hippocampal activity in older adults is comparable to young, consistent with their increased recognition accuracy under conditions of encoding support.

New method of blurred radius detection in defocused image

An algorithm combining local entropy and histogram statistics was proposed in order to solve the problem that it is difficult to detect the blurred radius quickly and correctly in defocused image blind restoration.Firstly,the gray level information of blurred image was extracted by local entropy filter and straight lines were detected by Canny edge detector and Hough transform.Secondly,the straight step edges were located to compute the line spread function adopting Grubbs method and histogram feature of the parallel line area.Finally,the blurred radius could be obtained by line spread function.The experimental results show that the step edges can be located quickly and accurately when blurred radius is small,thus increasing recognition accuracy and recognition efficiency.

Fast neuromimetic object recognition using FPGA outperforms GPU implementations.

Recognition of objects in still images has traditionally been regarded as a difficult computational problem. Although modern automated methods for visual object recognition have achieved steadily increasing recognition accuracy, even the most advanced computational vision approaches are unable to obtain performance equal to that of humans. This has led to the creation of many biologically inspired models of visual object recognition, among them the hierarchical model and X (HMAX) model. HMAX is traditionally known to achieve high accuracy in visual object recognition tasks at the expense of significant computational complexity. Increasing complexity, in turn, increases computation time, reducing the number of images that can be processed per unit time. In this paper we describe how the computationally intensive and biologically inspired HMAX model for visual object recognition can be modified for implementation on a commercial field-programmable aate Array, specifically the Xilinx Virtex 6 ML605 evaluation board with XC6VLX240T FPGA. We show that with minor modifications to the traditional HMAX model we can perform recognition on images of size 128 × 128 pixels at a rate of 190 images per second with a less than 1% loss in recognition accuracy in both binary and multiclass visual object recognition tasks.

Locating and recognizing multiple human actions by searching for maximum score subsequences

Despite the numerous methods to recognize human actions in a video, few are designed for videos containing more than one action over a certain time period. Moreover, existing multiple action recognition methods adopt windowed sequence search strategy. Windowed sequence searching requires an exhaustive trial of window length yielding intensive computation. This work presents a frame-based strategy, capable of searching for maximum score subsequences that correspond to actions. Therefore, start and ending times of all actions are located, and action categories are identified as well. Moreover, contrast mutual information is proposed as a new score function to increase recognition accuracy. Experimental results indicate that the proposed method locates and recognizes multiple actions in a video accurately, even for the conventional single action classification problem.

Pairwise constraint-guided sparsity preserving projections

Concerning the deficiency of supervision information in the process of sparse reconstruction in Sparsity Preserving Projection(SPP),Pairwise Constraint-guided Sparsity Preserving Projection(PCSPP) was proposed,which introduced supervision information of must-link constraints and cannot-link constraints to guide sparse reconstruction in the process of sparsity reconstruction of training samples,making SPP fuse constraint supervise information efficiently.The experimental results in UMIST,YALE and AR face datasets show,in contrast to unsupervised sparsity preserving projections,our algorithm achieves approximately 5%~15% increase in recognition accuracy based on the nearest neighbor classifier and promotes efficiently the performance of dimensionality reduction classification.

System-on-programmable-chip implementation of diminishing learning based pattern recognition system

Support vector machine (SVM) is one of the state-of-the-art tools for linear and nonlinear pattern classification. One of the design issues in SVM classifier is reducing the number of support vectors without compromising the classification accuracy. A technique denoted as diminishing learning (DL) is already proposed in literature for an SVM based multi-class isolated digit recognition system using speaker dependent TI46 database of isolated digits. In this paper, the computational complexity for SVM and SVM-DL based isolated digit recognition system is studied and the computation time for both the classifiers is evaluated by system-on-programmable-chip (SOPC) implementation of the recognition system onto an Altera Cyclone II Series FPGA using Nios II Soft-core processor. The number of support vectors is reduced by 38.28–90.25 % on using SVM-DL for isolated digit recognition problem. This in turn reduces the classification time for SVM-DL by 31.45–91.78 % over SVM. Recognition accuracies of 97 and 98 % are achieved for SVM classifier with and without DL technique, respectively. The study confirms the effect of, the order in which the classes are classified, on the recognition accuracy. For the TI46 database, about 2–4 % increase in recognition accuracy is obtained by choosing the optimum order for SVM-DL classifier. The proposed SOPC implementation of SVM-DL based recognition system can be employed for various other pattern recognition applications too such as face recognition, character recognition and target recognition.

Fuzzy Rule-Based System for Decision Making Support of Hybrid SVM-GMM Acoustic Event Detection

This paper presents a hybrid SVM-GMM mechanism based on a fuzzy rule-based system (FRBS), called FRBS-SVMGMM, for acoustic event detection (AED) applications. This method effectively combines the results of support vector machine (SVM) and Gaussian mixture model (GMM) calculations on acoustic data within the framework of FRBS operations. With the support of FRBS, which greatly increases AED recognition accuracy, AED can make reliable and correct detection decisions. This approach is essential to applications where low false recognition is a major concern. The proposed FRBS-SVMGMM mechanism is conceptually simple and computationally inexpensive. The comparative recognition performance experiments in this study demonstrate the effectiveness and superiority of the proposed FRBS-SVMGMM.

Genetic and evolutionary methods for biometric feature reduction

In this paper, we investigate the use of Genetic and Evolutionary Computations (GECs) for feature selection (GEFeS), weighting (GEFeW), and hybrid weighting/selection (GEFeWS) in an attempt to increase recognition accuracy as well as reduce the number of features needed for biometric recognition. These GEC-based methods were first applied to a subset of 105 subjects taken from the Facial Recognition Grand Challenge (FRGC) dataset (FRGC-105) where several feature masks were evolved. The resulting feature masks were then tested on a larger subset taken from the FRGC dataset (FRGC-209) in an effort to investigate how well they generalise to unseen subjects. The results suggest that our GEC-based methods are effective in increasing the recognition accuracy and reducing the features needed for recognition. In addition, the evolved FRGC-105 feature masks generalised well on the unseen subjects within the FRGC-209 dataset.

Efficient method of person authentication based on fusion of best bits in left and right irises

Multi-unit iris authentication system is proposed by deploying only the best bits in iris code. This reduced feature set provides better recognition rates in lesser time compared to other methods that use all the bits in iris code. To increase recognition accuracy further, our proposed method uses both irises of persons and performs score level fusion of distance values obtained from left and right irises using SVM, max, min and weighted sum methods. We have implemented this system using images from CASIA iris database. Experimental results show that fusion by SVM gives better recognition accuracy than other fusion rules.

Stochastic Subset Selection for Learning With Kernel Machines

Kernel machines have gained much popularity in applications of machine learning. Support vector machines (SVMs) are a subset of kernel machines and generalize well for classification, regression, and anomaly detection tasks. The training procedure for traditional SVMs involves solving a quadratic programming (QP) problem. The QP problem scales super linearly in computational effort with the number of training samples and is often used for the offline batch processing of data. Kernel machines operate by retaining a subset of observed data during training. The data vectors contained within this subset are referred to as support vectors (SVs). The work presented in this paper introduces a subset selection method for the use of kernel machines in online, changing environments. Our algorithm works by using a stochastic indexing technique when selecting a subset of SVs when computing the kernel expansion. The work described here is novel because it separates the selection of kernel basis functions from the training algorithm used. The subset selection algorithm presented here can be used in conjunction with any online training technique. It is important for online kernel machines to be computationally efficient due to the real-time requirements of online environments. Our algorithm is an important contribution because it scales linearly with the number of training samples and is compatible with current training techniques. Our algorithm outperforms standard techniques in terms of computational efficiency and provides increased recognition accuracy in our experiments. We provide results from experiments using both simulated and real-world data sets to verify our algorithm.

The potency of people in pictures: Evidence from sequences of eye fixations

Does the presence of people in a natural scene affect the way that we inspect that picture? Previous research suggests that we have a natural tendency to look at the social information before other items in a scene. There is also evidence that accuracy of visual memory and the way we move our eyes are related. This experiment investigated whether eye movements differed when participants correctly and incorrectly identified stimuli at recognition, and how this is affected by the presence of people. Eye movements were recorded from 15 participants while they inspected photographs at encoding and during a recognition memory test. Half of the pictures contained people and half did not. The presence of people increased recognition accuracy and affected average fixation duration and average saccadic amplitude. Accuracy was not affected by the size of the Region of Interest (RoI), the number of people in the picture, or the distance of the person from the center. Analyses of the order and pattern of fixations showed a high similarity between encoding and recognition in all conditions, but the lack of relationship between string similarity and recognition accuracy challenges the idea that the reproduction of eye movements alone is enough to create a memory advantage.

Visual exploration of emotional facial expressions in Parkinson's disease

Parkinson's disease (PD) is associated with impairments in facial emotion recognition as well as visual and executive dysfunction. We investigated whether facial emotion categorization impairments in PD are attributable to visual scanning abnormalities by recording the eye movements of 16 non-demented PD and 20 healthy control (HC) participants during an emotion recognition task. We examined the influence of several factors that can affect visual scanning, including oculomotor, basic visual, and cognitive abilities (executive function). Increases in the number and duration of fixations in the top regions of surprise facial expressions were related to increases in recognition accuracy for this emotion in PD participants with left-sided motor-symptom onset. Compared to HC men, HC women spent less time fixating on fearful expressions. PD participants displayed oculomotor abnormalities (antisaccades), but these were unrelated to scanning patterns. Performance on visual measures (acuity, contrast sensitivity) correlated with scanning patterns in the PD group only. Poorer executive function was associated with longer fixation times in PD and with a greater number of fixations in HC. Our findings indicate a specific relation between facial emotion categorization impairments and scanning of facial expressions in PD. Furthermore, PD and HC participants’ scanning behaviors during an emotion categorization task were driven by different perceptual processes and cognitive strategies. Our results underscore the need to consider differences in perceptual and cognitive abilities in studies of visual scanning, particularly when examining this ability in patient populations for which both vision and cognition are impaired.