Traffic Data Collection Systems: VEKPI - Count.

In recent years, a variety of supervised manifold learning techniques have been proposed to outperform their unsupervised alternative versions in terms of classification accuracy and data structure capturing. Some dissimilarity measures have been used in these techniques to guide the dimensionality reduction process. Their good performance was empirically demonstrated; however, the relevant analysis is still missing. This paper contributes to a theoretical analysis on a) how dissimilarity measures affect maintaining manifold neighbourhood structure and b) how supervised manifold learning techniques could contribute to the reduction of classification error. This paper also provides a cross-comparison between supervised and unsupervised manifold learning approaches in terms of structure capturing using Kendall's Tau coefficients and co-ranking matrices. Four different metrics (including three dissimilarity measures and Euclidean distance) have been considered along with manifold learning methods such as Isomap, t-Stochastic Neighbour Embedding ( t-SNE), and Laplacian Eigenmaps (LE), in two datasets: Breast Cancer and Swiss-Roll. This paper concludes that although the dissimilarity measures used in the manifold learning techniques can reduce classification error, they do not learn well or preserve the structure of the hidden manifold in the high dimensional space, but instead, they destroy the structure of the data. Based on the findings of this paper, it is advisable to use supervised manifold learning techniques as a pre-processing step in classification. In addition, it is not advisable to apply supervised manifold learning for visualization purposes since the two-dimensional representation using supervised manifold learning does not improve the preservation of data structure.

National roads are one of the main networks of a country’s transportation system. To maintain the performance level of national roads requires a well-structured pavement management system (PMS). The decision support system (DSS) is inseparable in the modern PMS, which required the development of a new approach for the DSS in support of national road network maintenance. The proposed model integrates data mining (DM) and geographical information system (GIS) to construct a simple DSS. DM is used to developed road maintenance optimization models, and then integrated with DSS with the help of GIS as an interface application. Historical data on the national road network in West Java, Indonesia is used as a case study. Examples show that the proposed model can determine a decision support solution efficiently. In addition, a userfriendly computer interface is developed so that PMS stakeholders can plan pavement maintenance simply and effectively.

The design of pattern recognition-based myoelectric interfaces has been heavily explored and contested in the research literature. A considerable proportion of the performance of these interfaces has been linked to the quality of the feature extraction (FE) stage used to describe the underlying myoelectric signal. In this paper, we address two important factors of FE that have not been fully exploited; 1) Spatial focus - traditional FE methods focus mainly on the concatenation of features extracted from individual channels and 2) Temporal focus - available FE methods are cross-sectional in nature, largely ignoring the temporal information that may exist between feature windows. To overcome these limitations, several spatiotemporal FE methods have been proposed including hand-crafted and deep learning (DL) models, with the latter showing significant performance enhancements at the cost of increased computational burden. This paper tackles the aforementioned limitations by 1) proposing novel extensions to simple time-domain FE methods, including the waveform length, zero crossings and root mean square, that can capture the relation between any number of channels, and 2) leveraging the long short-term memory concepts from deep neural networks to build a recursive framework that overcomes the cross-sectional nature of traditional methods. The advantages offered by the proposed Recursive Multi-Signal Temporal Fusion (RMTF) features include improved performance, competing with state-of-the-art FE methods without the computational costs associated with leading DL models, and the simplicity of the concepts making them suitable for real-time implementations. Experiments on 65 intact-limbed and amputee subjects reveal an approximate average of 15% reduction in classification errors as compared to models built with other feature sets.

This paper investigates the contribution of features which convey long-term spectro-temporal (ST) information for the purpose of automatic emotional speech classification. The ST representation is obtained by means of a modulation filterbank decomposition of long-term temporal envelopes of the outputs of a gammatone filterbank. The two-dimensional discrete cosine transform is used to reduce the dimensionality of the representation; candidate features are then derived from statistics computed from the DCT coefficients. Sequential forward feature selection is used to select the most salient features. Two types of experiments are described which use the Berlin emotional speech database to test the performance of the ST features alone and in combination with prosodic features. In a multi-class experiment, simulation results with a support vector classifier show that a 44% reduction in classification error is attained once prosodic features are combined with the proposed ST features. Additionally, in a one-against-all experiment, an average increase in F-score of 33% is attained when the proposed ST features are included.

Multiview, broadband, acoustic classification of individual fish was investigated using a recently developed laboratory scattering system. Scattering data from nine different species of saltwater fish were collected. Using custom software, these data were processed and filtered to yield a data set of 36 individuals, and between 200 and 500 echoes per individual. These data were sampled uniformly randomly in fish orientation. Feature-, decision-, and collaborative-fusion algorithms were then developed and tested using support vector machines (SVMs) as the underlying classifiers. Decision fusion was implemented by cascading two levels of support vectors machines. Collaborative fusion was implemented by using SVM outputs to estimate confidence levels and performing weighted averaging of probabilities computed from each view with feedback from other views. Collaborative fusion performed as well or better than the others, and did so without requiring assumptions about view geometry. In addition to a comparison between classification algorithms and feature transformations, two data collection geometries were explored, including random observation geometries. In all cases, combining multiple, broadband views yielded significant reductions in classification error (>;50%) over single-view methods, for uniformly random fish orientation.

On-implant spike sorting methods employ static feature extraction/selection techniques to minimize the hardware cost. Here we propose a novel framework for real-time spike sorting based on dynamic selection of features. We select salient features that maximize the geometric-mean of between-class distances as well as the associated homogeneity index effectively to best discriminate spikes for classification. Wave-shape classification is performed based on a multi-label window discrimination approach. An external module calculates the salient features and discrimination windows through optimizing a replica of the on-implant operation, and then configures the on-implant spike sorter for real-time online operation. Hardware implementation of the on-implant online spike sorter for 512 channels of concurrent extra-cellular neural signals is reported, with an average classification accuracy of ~88%. Compared with other similar methods, our method shows reduction in classification error by a factor of ~2, and also reduction in the required memory space by a factor of ~5.

This project includes two main parts: Development of a “Selective Wavelength Imaging Sensor” and an “Adaptive Classifiery System” for adaptive imaging and sorting of agricultural products respectively. Three different technologies were investigated for building a selectable wavelength imaging sensor: diffraction gratings, tunable filters and linear variable filters. Each technology was analyzed and evaluated as the basis for implementing the adaptive sensor. Acousto optic tunable filters were found to be most suitable for the selective wavelength imaging sensor. Consequently, a selectable wavelength imaging sensor was constructed and tested using the selected technology. The sensor was tested and algorithms for multispectral image acquisition were developed. A high speed inspection system for fresh-market carrots was built and tested. It was shown that a combination of efficient parallel processing of a DSP and a PC based host CPU in conjunction with a hierarchical classification system, yielded an inspection system capable of handling 2 carrots per second with a classification accuracy of more than 90%. The adaptive sorting technique was extensively investigated and conclusively demonstrated to reduce misclassification rates in comparison to conventional non-adaptive sorting. The adaptive classifier algorithm was modeled and reduced to a series of modules that can be added to any existing produce sorting machine. A simulation of the entire process was created in Matlab using a graphical user interface technique to promote the accessibility of the difficult theoretical subjects. Typical Grade classifiers based on k-Nearest Neighbor techniques and linear discriminants were implemented. The sample histogram, estimating the cumulative distribution function (CDF), was chosen as a characterizing feature of prototype populations, whereby the Kolmogorov-Smirnov statistic was employed as a population classifier. Simulations were run on artificial data with two-dimensions, four populations and three classes. A quantitative analysis of the adaptive classifier's dependence on population separation, training set size, and stack length determined optimal values for the different parameters involved. The technique was also applied to a real produce sorting problem, e.g. an automatic machine for sorting dates by machine vision in an Israeli date packinghouse. Extensive simulations were run on actual sorting data of dates collected over a 4 month period. In all cases, the results showed a clear reduction in classification error by using the adaptive technique versus non-adaptive sorting.

Active reducing classification error for CAD systems

Many variants of local binary patterns (LBPs) are widely used for face analysis due to their inherent simplicity and robustness. However, it has not yet been proven that LBPs are optimal for this task in regards to achieving the best balance between minimizing code numbers and reducing classification error. We propose an effective code selection method for selecting optimal LBP (OLBP) based on the maximization of mutual information (MMI) between features and class labels. We demonstrate the effectiveness of the proposed OLBP through several experiments of facial expression recognition. Experimental results show that the OLBP outperforms other features such as LBP, ULBP, and MCT in terms of minimizing the number of codes and reducing the classification error.

In the present context, it is frequent for urban and national road networks to be highly congested, resulting in increased travel times and delays. At the same time, due to toll charging, some high-quality highway networks, particularly in Portugal, are underused. Dynamic charging on highway networks can contribute towards the optimization of the network performance and understanding the behavior of logistic operators as one of the main users of the network is of great importance for the development of any dynamic charging toll model. Within this context, one of the pilot cases of the H2020 project OPTIMUM includes the development of such a forecasting and dynamic (toll) charging model with the key objective of transferring heavy traffic from the urban and national roads to highways. The first run of pilot occurred between March 09th, 2017 and April 07th, 2017, performance indicators (KPI) were highly positive in terms of freight traffic shift from the national roads to the highways and, consequently, an increase in toll revenues and freight operations efficiency was registered.

This paper presents the results of an advanced inspection, structural assessment, and development of rehabilitation and strengthening strategies for Concrete Bridge No. 44, located on National Road N2, Prishtina - Blacë. The primary objective of this study is to identify effective methods for restoring load-bearing capacity and improving the operational safety of the bridge, in compliance with contemporary design standards as prescribed by the Eurocodes, considering the current degradation state of its structural elements and the critical importance of this infrastructure within the national road network. The assessment methodology involved systematic visual inspections, non-destructive testing, and numerically modeled structural analyses to evaluate the existing performance of the main load-bearing beams, columns, and other key components. Based on the collected data, modular and phased intervention proposals have been developed, allowing for the prioritized implementation of repair measures according to the severity of identified deficiencies and the functional priorities of the bridge. The results of this study provide both practical and scientific contributions to the planning of interventions on existing structures built in previous decades, combining modern rehabilitation engineering approaches with international standards of structural design and safety control.

Tree-based models are popular and widely used because they are simple, flexible and powerful tools for classification. Unfortunately they are not stable classifiers. Significant improvement of model stability and prediction accuracy can be obtained by aggregation of multiple classification trees. The reduction of classification error is a result of decreasing bias or/and variance of the committee of trees (called also an ensemble or a forest). In this paper we discuss and compare different methods for model aggregation. We also address the problem of finding minimal number of trees sufficient for the forest.

National and international road networks try to grow in efficiency and safety with the support of ITS (Intelligent Transportation Systems). Focused on every intricacy of such sophisticated and massive communication systems, road operators and managers risk disregarding the most important step: effective communication with national and international road users. The Mare Nostrum VMS Group intends to improve harmonization of VMS displays all through a Long Distance Corridor that spans from Seville (Spain) to Trieste (Italy). The group is currently searching for international abstract symbols to indicate distance, length, and its combinations. Following the comprehension judgment method, this paper presents some empirical results aimed at comparing the current signing solutions, and filtering out the worse alternatives.

This paper presents a novel fusion technique for automatic target recognition from high range resolution radar profiles when observations from multiple viewpoints are available. The fusion technique entails only a straightforward modification of the transition probabilities of a single-viewpoint target model in which a Hidden Markov Model is used to represent the unknown target orientation. Evaluations using the MSTAR database indicate that the new technique can reduce classification errors by about two orders of magnitude when compared to single viewpoint observations and, in a 10-target classification experiment, gave almost perfect recognition.

ROLEX-SP: Rules of lexical syntactic patterns for free text categorization