IEEE Conf Research Articles

Convolutional Neural Networks Quantization with Double-Stage Squeeze-and-Threshold.

It has been proven that, compared to using 32-bit floating-point numbers in the training phase, Deep Convolutional Neural Networks (DCNNs) can operate with low-precision during inference, thereby saving memory footprint and power consumption. However, neural network quantization is always accompanied by accuracy degradation. Here, we propose a quantization method called double-stage Squeeze-and-Threshold (double-stage ST) to close the accuracy gap with full-precision models. While accurate colors in pictures can be pleasing to the viewer, they are not necessary for distinguishing objects. The era of black and white television proves this idea. As long as the limited colors are filled reasonably for different objects, the objects can be well identified and distinguished. Our method utilizes the attention mechanism to adjust the activations and learn the thresholds to distinguish objects (features). We then divide the numerically rich activations into intervals (a limited variety of numerical values) by the learned thresholds. The proposed method supports both binarization and multi-bit quantization. Our method achieves state-of-the-art results. In binarization, ReActNet [Z. Liu, Z. Shen, S. Li, K. Helwegen, D. Huang and K. Cheng, arXiv:abs/2106.11309] trained with our method outperforms the previous state-of-the-art result by 0.2 percentage points. Whereas in multi-bit quantization, the top-1 accuracy of the 3-bit ResNet-18 [K. He, X. Zhang, S. Ren and J. Sun, Deep residual learning for image recognition, 2016 IEEE Conf. Computer Vision and Pattern Recognition, CVPR 2016, 27-30 June 2016, Las Vegas, NV, USA (IEEE Computer Society, 2016), pp. 770-778] model exceeds the top-1 accuracy of its full-precision baseline model by 0.4 percentage points. The double-stage ST activation quantization method is easy to apply by inserting it before the convolution. Besides, the double-stage ST is detachable after training and introducing no computational cost in inference.

Robust Person Re-Identification Through the Combination of Metric Learning and Late Fusion Techniques

Fusion techniques with the aim to leverage the discriminative power of different appearance features for person representation have been widely applied in person re-identification. They are performed by concatenating all feature vectors (known as early fusion) or by combining matching scores of different classifiers (known as late fusion). Previous studies have proved that late fusion techniques achieve better results than early fusion ones. However, majority of the studies focus on determining the suitable weighting schemes that can reflect the role of each feature. The determined weights are then integrated in conventional similarity functions, such as Cosine [L. Zheng, S. Wang, L. Tian, F. He, Z. Liu and Q. Tian, Queryadaptive late fusion for image search and person reidentification, in Proc. IEEE Conf. Computer Vision and Pattern Recognition, 2015, pp. 1741–1750]. The contribution of this paper is two-fold. First, a robust person re-identification method by combining the metric learning with late fusion techniques is proposed. The metric learning method Cross-view Quadratic Discriminant Analysis (XQDA) is employed to learn a discriminant low dimensional subspace to minimize the intra-person distance while maximize the inter-person distance. Moreover, product rule-based and sum rule-based late fusion techniques are applied on these distances. Second, concerning feature engineering, the ResNet extraction process has been modified in order to extract local features of different stripes in person images. To show the effectiveness of the proposed method, both single-shot and multi-shot scenarios are considered. Three state-of-the-art features that are Gaussians of Gaussians (GOG), Local Maximal Occurrence (LOMO) and deep-learned features extracted through a Residual network (ResNet) are extracted from person images. The experimental results on three benchmark datasets that are iLIDS-VID, PRID-2011 and VIPeR show that the proposed method [Formula: see text]% [Formula: see text]% of improvement over the best results obtained with the single feature. The proposed method that achieves the accuracy of 85.73%, 93.82% and 50.85% at rank-1 for iLIDS-VID, PRID-2011 and VIPeR, respectively, outperforms different SOTA methods including deep learning ones. Source code is publicly available to facilitate the development of person re-ID system.

Self-Supervised Multi-View Person Association and its Applications.

Reliable markerless motion tracking of people participating in a complex group activity from multiple moving cameras is challenging due to frequent occlusions, strong viewpoint and appearance variations, and asynchronous video streams. To solve this problem, reliable association of the same person across distant viewpoints and temporal instances is essential. We present a self-supervised framework to adapt a generic person appearance descriptor to the unlabeled videos by exploiting motion tracking, mutual exclusion constraints, and multi-view geometry. The adapted discriminative descriptor is used in a tracking-by-clustering formulation. We validate the effectiveness of our descriptor learning on WILDTRACK T. Chavdarova et al., "WILDTRACK: A multi-camera HD dataset for dense unscripted pedestrian detection," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2018, pp. 5030-5039. and three new complex social scenes captured by multiple cameras with up to 60 people "in the wild". We report significant improvement in association accuracy (up to 18 percent) and stable and coherent 3D human skeleton tracking (5 to 10 times) over the baseline. Using the reconstructed 3D skeletons, we cut the input videos into a multi-angle video where the image of a specified person is shown from the best visible front-facing camera. Our algorithm detects inter-human occlusion to determine the camera switching moment while still maintaining the flow of the action well. Website: http://www.cs.cmu.edu/~ILIM/projects/IM/Association4Tracking.

Enhancing Blurred Low-Resolution Images via Exploring the Potentials of Learning-Based Super-Resolution

This paper aims to propose a candidate solution to the challenging task of single-image blind super-resolution (SR), via extensively exploring the potentials of learning-based SR schemes in the literature. The task is formulated into an energy functional to be minimized with respect to both an intermediate super-resolved image and a nonparametric blur-kernel. The functional includes a so-called convolutional consistency term which incorporates a nonblind learning-based SR result to better guide the kernel estimation process, and a bi-[Formula: see text]-[Formula: see text]-norm regularization imposed on both the super-resolved sharp image and the nonparametric blur-kernel. A numerical algorithm is deduced via coupling the splitting augmented Lagrangian (SAL) and the conjugate gradient (CG) method. With the estimated blur-kernel, the final SR image is reconstructed using a simple TV-based nonblind SR method. The proposed blind SR approach is demonstrated to achieve better performance than [T. Michaeli and M. Irani, Nonparametric Blind Super-resolution, in Proc. IEEE Conf. Comput. Vision (IEEE Press, Washington, 2013), pp. 945–952.] in terms of both blur-kernel estimation accuracy and image ehancement quality. In the meanwhile, the experimental results demonstrate surprisingly that the local linear regression-based SR method, anchored neighbor regression (ANR) serves the proposed functional more appropriately than those harnessing the deep convolutional neural networks.

Incorporating Trust and Self-Confidence Analysis in the Guidance and Control of (Semi)Autonomous Mobile Robotic Systems

We propose a trust and self-confidence-based autonomy allocation strategy to automatically choose between manual and autonomous control of (semi)autonomous mobile robots in guidance and navigation tasks. We utilize a performance-centric, computational trust and self-confidence model and automated autonomy allocation strategy, developed in our earlier work (H. Saeidi and Y. Wang, “Trust and self-confidence based autonomy allocation for robotic systems,” in Proc. 54th IEEE Conf. Decis. Control , 2015, pp. 6052–6057.), based on objective and unbiased performance measures for the human and the robot. A set of robot simulations with a human-in-the-loop is conducted for a teleoperated unmanned aerial vehicle tracking task. The results demonstrate that our allocation strategy can capture human autonomy allocation pattern with an accuracy of 64.05%. We also show that the strategy can improve the overall robot performance by 11.76% and reduce operator's workload by 10.07% compared to a manual allocation. Moreover, compared to a performance maximization strategy, our strategy is 23.42% more likely to be accepted and generally preferred and trusted by the participants. Furthermore, we design a decision pattern correction algorithm based on nonlinear model predictive control to help a human operator gradually adapt to a modified allocation pattern for improved overall performance.

An Illumination-Invariant Nonparametric Model for Urban Road Detection

In this paper, we propose an illumination-invariant nonparametric model for urban road detection based on a monocular camera and a single-line LIDAR sensor. With the monocular camera, we propose a new shadow removal method to obtain an illumination-invariant image representation. Consequently, we can accurately locate the road vanishing point after removing the adverse shadowy effect. With the constraint of the detected vanishing point, we propose a Dijkstra-based method to compute a minimum-cost map, where the minimum-cost path from the vanishing point to any other pixel can be found. With the single line LIDAR sensor, we can locate a few potential curb points in the image bottom region, and thus we can obtain several corresponding minimum-cost paths that originate from the vanishing point to the curb points. Thereafter, two most likely road borders can be found from these paths, respectively. Our learning-free method has been tested on over 4000 images of the KITTI-Odometry Dataset [A. Geiger, P. Lenz, and R.Urtasun, “Are we ready for autonomous driving? The KITTI vision benchmark suite,” in Proc. IEEE Conf. Comput. Vision Pattern Recognit., 2012, pp. 3354-3361.] and the Oxford Robotcar Dataset [W. Maddern, G. Pascoe, C. Linegar, and P. Newman, “1 year, 1000 km: The Oxford robotcar dataset,” Int. J. Robot. Res., vol. 36, no. 1, pp. 3-15, 2017.]. It works accurately on a variety of road scenes and is competitive compared to state-of-the-art deep learning methods that need extensive training data.

Guest Editorial Special Issue of Best Conference Papers in IEEE Conf. in Latin America 2017

Guest Editorial Special Issue of Best Conference Papers in IEEE Conf. in Latin America 2017

Witnessing matrix identities and proof complexity

We use results from the theory of algebras with polynomial identities (PI-algebras) to study the witness complexity of matrix identities. A matrix identity of [Formula: see text] matrices over a field [Formula: see text]is a non-commutative polynomial (f(x1, …, xn)) over [Formula: see text], such that [Formula: see text] vanishes on every [Formula: see text] matrix assignment to its variables. For every field [Formula: see text]of characteristic 0, every [Formula: see text] and every finite basis of [Formula: see text] matrix identities over [Formula: see text], we show there exists a family of matrix identities [Formula: see text], such that each [Formula: see text] has [Formula: see text] variables and requires at least [Formula: see text] many generators to generate, where the generators are substitution instances of elements from the basis. The lower bound argument uses fundamental results from PI-algebras together with a generalization of the arguments in [P. Hrubeš, How much commutativity is needed to prove polynomial identities? Electronic colloquium on computational complexity, ECCC, Report No.: TR11-088, June 2011].We apply this result in algebraic proof complexity, focusing on proof systems for polynomial identities (PI proofs) which operate with algebraic circuits and whose axioms are the polynomial-ring axioms [P. Hrubeš and I. Tzameret, The proof complexity of polynomial identities, in Proc. 24th Annual IEEE Conf. Computational Complexity, CCC 2009, 15–18 July 2009, Paris, France (2009), pp. 41–51; Short proofs for the determinant identities, SIAM J. Comput. 44(2) (2015) 340–383], and their subsystems. We identify a decrease in strength hierarchy of subsystems of PI proofs, in which the [Formula: see text]th level is a sound and complete proof system for proving [Formula: see text] matrix identities (over a given field). For each level [Formula: see text] in the hierarchy, we establish an [Formula: see text] lower bound on the number of proof-steps needed to prove certain identities.Finally, we present several concrete open problems about non-commutative algebraic circuits and speed-ups in proof complexity, whose solution would establish stronger size lower bounds on PI proofs of matrix identities, and beyond.

Continuous Motion Classification and Segmentation Based on Improved Dynamic Time Warping Algorithm

The continuous image sequence recognition is more difficult than the single image recognition because the classification of continuous image sequences and the image edge recognition must be very accurate. Hence, a method based on sequence alignment for action segmentation and classification is proposed to reconstruct a template sequence by estimating the mean action of a class category, which calculates the distance between a single image and a template sequence by sparse coding in Dynamic Time Warping. The proposed method, the methods of Kulkarni et al. [Continuous action recognition based on sequence alignment, Int. J. Comput. Vis. pp. 1–26.] and Hoai et al. [Joint segmentation and classification of human actions in video, IEEE Conf. Computer Vision and Pattern Recognition, 2008, pp. 108–119.] are compared on the recognition accuracy of the continuous recognition and isolated recognition, which clearly shows that the proposed method outperforms the other methods. When applied to continuous gesture classification, it not only can recognize the gesture categories more quickly and accurately, but is more realistic in solving continuous action recognition problems in a video than the other existing methods.

Likelihood Analysis of Power Spectra and Generalized Moment Problems

We develop an approach to the spectral estimation that has been advocated by [A. Ferrante et al. , “Time and spectral domain relative entropy: A new approach to multivariate spectral estimation,” IEEE Trans. Autom. Control , vol. 57, no. 10, pp. 2561–2575, Oct. 2012.] and, in the context of the scalar-valued covariance extension problem, by [P. Enqvist and J. Karlsson, “Minimal itakura-saito distance and covariance interpolation,” in Proc. 47th IEEE Conf. Decision Control , 2008, pp. 137–142]. The aim is to determine the power spectrum that is consistent with given moments and minimizes the relative entropy between the probability law of the underlying Gaussian stochastic process to that of a prior. The approach is analogous to the framework of earlier work by Byrnes, Georgiou, and Lindquist and can also be viewed as a generalization of the classical work by Burg and Jaynes on the maximum entropy method. In this paper, we present a new fast algorithm in the general case (i.e., for general Gaussian priors) and show that for priors with a specific structure the solution can be given in closed form.

Scale-aware Fast R-CNN for Pedestrian Detection

In this paper, we consider the problem of pedestrian detection in natural scenes. Intuitively, instances of pedestrians with different spatial scales may exhibit dramatically different features. Thus, large variance in instance scales, which results in undesirable large intracategory variance in features, may severely hurt the performance of modern object instance detection methods. We argue that this issue can be substantially alleviated by the divide-and-conquer philosophy. Taking pedestrian detection as an example, we illustrate how we can leverage this philosophy to develop a Scale-Aware Fast R-CNN (SAF R-CNN) framework. The model introduces multiple built-in subnetworks which detect pedestrians with scales from disjoint ranges. Outputs from all of the subnetworks are then adaptively combined to generate the final detection results that are shown to be robust to large variance in instance scales, via a gate function defined over the sizes of object proposals. Extensive evaluations on several challenging pedestrian detection datasets well demonstrate the effectiveness of the proposed SAF R-CNN. Particularly, our method achieves state-of-the-art performance on Caltech [P. Dollar, C. Wojek, B. Schiele, and P. Perona, “Pedestrian detection: An evaluation of the state of the art,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 34, no. 4, pp. 743–761, Apr. 2012], and obtains competitive results on INRIA [N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2005, pp. 886–893], ETH [A. Ess, B. Leibe, and L. V. Gool, “Depth and appearance for mobile scene analysis,” in Proc. Int. Conf. Comput. Vis ., 2007, pp. 1–8], and KITTI [A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? The KITTI vision benchmark suite,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit ., 2012, pp. 3354–3361].

Fisher Motion Descriptor for Multiview Gait Recognition

The goal of this paper is to identify individuals by analyzing their gait. Instead of using binary silhouettes as input data (as done in many previous works) we propose and evaluate the use of motion descriptors based on densely sampled short-term trajectories. We take advantage of state-of-the-art people detectors to define custom spatial configurations of the descriptors around the target person, obtaining a rich representation of the gait motion. The local motion features (described by the Divergence-Curl-Shear descriptor [M. Jain, H. Jegou and P. Bouthemy, Better exploiting motion for better action recognition, in Proc. IEEE Conf. Computer Vision Pattern Recognition (CVPR) (2013), pp. 2555–2562.]) extracted on the different spatial areas of the person are combined into a single high-level gait descriptor by using the Fisher Vector encoding [F. Perronnin, J. Sánchez and T. Mensink, Improving the Fisher kernel for large-scale image classification, in Proc. European Conf. Computer Vision (ECCV) (2010), pp. 143–156]. The proposed approach, coined Pyramidal Fisher Motion, is experimentally validated on ‘CASIA’ dataset [S. Yu, D. Tan and T. Tan, A framework for evaluating the effect of view angle, clothing and carrying condition on gait recognition, in Proc. Int. Conf. Pattern Recognition, Vol. 4 (2006), pp. 441–444]. (parts B and C), ‘TUM GAID’ dataset, [M. Hofmann, J. Geiger, S. Bachmann, B. Schuller and G. Rigoll, The TUM Gait from Audio, Image and Depth (GAID) database: Multimodal recognition of subjects and traits, J. Vis. Commun. Image Represent. 25(1) (2014) 195–206]. ‘CMU MoBo’ dataset [R. Gross and J. Shi, The CMU Motion of Body (MoBo) database, Technical Report CMU-RI-TR-01-18, Robotics Institute (2001)]. and the recent ‘AVA Multiview Gait’ dataset [D. López-Fernández, F. Madrid-Cuevas, A. Carmona-Poyato, M. Marín-Jiménez and R. Muñoz-Salinas, The AVA multi-view dataset for gait recognition, in Activity Monitoring by Multiple Distributed Sensing, Lecture Notes in Computer Science (Springer, 2014), pp. 26–39]. The results show that this new approach achieves state-of-the-art results in the problem of gait recognition, allowing to recognize walking people from diverse viewpoints on single and multiple camera setups, wearing different clothes, carrying bags, walking at diverse speeds and not limited to straight walking paths.

2015 IEEE Conf on Electron Devices and Solid-State Circuits

2015 IEEE Conf on Electron Devices and Solid-State Circuits

Visual Analysis of Social Media Data

The application of visual analytics, which combines the advantages of computational knowledge discovery and interactive visualization, to social media data highlights the many benefits of this integrated approach. The Web extra at http://youtu.be/nhoq71gqyXE is a video demonstrating a prototype system for visual-interactive analysis of large georeferenced microblog datasets, describing the design of the system, and detailing its application to the VAST 2011 Challenge dataset. The dataset models an epidemic outbreak in a fictitious metropolitan area. The video shows how the system can detect the epidemic and analyze its development over time. The system was implemented by Juri Buchmueller, Fabian Maass, Stephan Sellien, Florian Stoffel, and Matthias Zieker at the University of Konstanz (they also produced this video). Further information on the system and the VAST challenge dataset can be found in E. Bertini et al., Analytics of Terrorist Activities Related to Epidemics, Proc. IEEE Conf. Visual Analytics Science and Technology (VAST 11), IEEE CS, pp. 329-330, 2011.

A Kharitonov-like theorem for robust stability independent of delay of interval quasipolynomials

In this paper, a Kharitonov-like theorem is proved for testing robust stability independent of delay of interval quasipolynomials, p(s)+∑eqk(s), where p and qk's are interval polynomials with uncertain coefficients. It is shown that the robust stability test of the quasipolynomial basically reduces to the stability test of a set of Kharitonov-like vertex quasipolynomials, where stability is interpreted as stability independent of delay. As discovered in (IEEE Trans. Autom. Control 2008; 53:1219–1234), the well-known vertex-type robust stability result reported in (IMA J. Math. Contr. Info. 1988; 5:117–123) (See also (IEEE Trans. Circ. Syst. 1990; 37(7):969–972; Proc. 34th IEEE Conf. Decision Contr., New Orleans, LA, December 1995; 392–394) does contain a flaw. An alternative approach is proposed in (IEEE Trans. Autom. Control 2008; 53:1219–1234), and both frequency sweeping and vertex type robust stability tests are developed for quasipolynomials with polytopic coefficient uncertainties. Under a specific assumption, it is shown in (IEEE Trans. Autom. Control 2008; 53:1219–1234) that robust stability independent of delay of an interval quasipolynomial can be reduced to stability independent of delay of a set of Kharitonov-like vertex quasipolynomials. In this paper, we show that the assumption made in (IEEE Trans. Autom. Control 2008; 53:1219–1234) is redundant, and the Kharitonov-like result reported in (IEEE Trans. Autom. Control 2008; 53:1219–1234) is true without any additional assumption, and can be applied to all quasipolynomials. The key idea used in (IEEE Trans. Autom. Control 2008; 53:1219–1234) was the equivalence of Hurwitz stability and ℂ-o-stability for interval polynomials with constant term never equal to zero. This simple observation implies that the well-known Kharitonov theorem for Hurwitz stability can be applied for ℂ-o-stability, provided that the constant term of the interval polynomial never vanishes. However, this line of approach is based on a specific assumption, which we call the CNF-assumption. In this paper, we follow a different approach: First, robust ℂ-o-stability problem is studied in a more general framework, including the cases where degree drop is allowed, and the constant term as well as other higher-orders terms can vanish. Then, generalized Kharitonov-like theorems are proved for ℂ-o-stability, and inspired by the techniques used in (IEEE Trans. Autom. Control 2008; 53:1219–1234), it is shown that robust stability independent of delay of an interval quasipolynomial can be reduced to stability independent of delay of a set of Kharitonov-like vertex quasipolynomials, even if the assumption adopted in (IEEE Trans. Autom. Control 2008; 53:1219–1234) is not satisfied. Copyright © 2009 John Wiley & Sons, Ltd.

Autonomous rigid body attitude synchronization

Control laws to synchronize attitudes in a swarm of fully actuated rigid bodies, in the absence of a common reference attitude or hierarchy in the swarm, are proposed in [Smith, T. R., Hanssmann, H., & Leonard, N.E. (2001). Orientation control of multiple underwater vehicles with symmetry-breaking potentials. In Proc. 40th IEEE conf. decision and control (pp. 4598–4603); Nair, S., Leonard, N. E. (2007). Stable synchronization of rigid body networks. Networks and Heterogeneous Media, 2(4), 595–624]. The present paper studies two separate extensions with the same energy shaping approach: (i) locally synchronizing the rigid bodies’ attitudes, but without restricting their final motion and (ii) relaxing the communication topology from undirected, fixed and connected to directed, varying and uniformly connected. The specific strategies that must be developed for these extensions illustrate the limitations of attitude control with reduced information.

Non-rigid object tracking in complex scenes

A recently proposed colour based tracking algorithm has been established to track objects in real circumstances [Zivkovic, Z., Kröse, B. 2004. An EM-like algorithm for color-histogram-based object tracking. In: Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 798–803]. To improve the performance of this technique in complex scenes, in this paper we propose a new algorithm for optimally adapting the ellipse outlining the objects of interest. This paper presents a Lagrangian based method to integrate a regularising component into the covariance matrix to be computed. Technically, we intend to reduce the residuals between the estimated probability distribution and the expected one. We argue that, by doing this, the shape of the ellipse can be properly adapted in the tracking stage. Experimental results show that the proposed method has favourable performance in shape adaption and object localisation.

Discrete-Time Expectation Maximization Algorithms for Markov-Modulated Poisson Processes

In this paper, we consider parameter estimation Markov-modulated Poisson processes via robust filtering and smoothing techniques. Using the expectation maximization algorithm framework, our filters and smoothers can be applied to estimate the parameters of our model in either an online configuration or an offline configuration. Further, our estimator dynamics do not involve stochastic integrals and our new formulas, in terms of time integrals, are easily discretized, and are written in numerically stable forms in W. P. Malcolm, R. J. Elliott, and J. van der Hoek, ldquoOn the numerical stability of time-discretized state estimation via clark transformations,rdquo presented at the IEEE Conf. Decision Control, Mauii, HI, Dec. 2003.

Driver steering assistance to avoid unintended lane departure by lane keeping and steering suggestions

One of the most frequent causes of highway accidents is the unintended out of lane deviation, due to moments of inattention, sleepiness or drowsiness. The enclosed video shows a driver steering assistance that prevents this type of lane departure. For vehicles with conventional steering columns, there are some difficulties in carrying out this type of steering assistance, in closed loop control. These difficulties consist in sharing vehicle control with the driver. The solution presented in this contribution is based on alternating the steering control of the vehicle between the driver and the steering assistance [1].The activation of the steering assistance takes place when the driver has a moment of inattention and the vehicle is going off the lane.• The driver's attention has been taken into account using the driver's torque on the steering wheel. If the driver's steering torque is below a certain threshold, then he is considered to be inattentive. Other driver monitoring systems could be used instead of the driver torque, without changing anything to the assistance.• The vehicle is considered to be going off the lane, when its front wheels have exceeded a central lane strip, called “normal driving” zone.While active, the steering assistance uses an electrical DC motor installed on the steering column. The DC motor provides a steering torque that brings the vehicle back to the centre of the lane and maintains it there until the driver recovers attention.Therefore, the steering torque follows an optimized control law. To design this control law, linear convex optimization methods (LMI) have been used. Three objectives have been taken into account.1. The first objective is to achieve a bounded overshoot of the front of the vehicle with respect to the considered “normal driving” zone.2. The second objective is for the entire vehicle state vector (side slip angle of the tires, yaw rate, relative yaw angle, lateral offset, steering angle and steering angle rate) to be bounded to safe values during the control.3. The third objective is to ensure a smooth response of the steering assistance after the activation, by using a bounded torque input from the DC motor.The deactivation of the steering assistance can take place under two circumstances.1. One is when the driver has recovered attention under the condition that the vehicle had already been driven back to the “normal driving” zone. This activation/deactivation strategy has been chosen considering the switched system: vehicle driven by its driver / vehicle driven by the steering assistance. The switching rules preserve the state boundedness of the switched system through switching in an invariant set.2. The other type of deactivation allows the driver to immediately switch off the system, for safety reasons, by counteracting the assistance, even if the vehicle is outside the “normal driving” zone. In this case, no guarantee for the state boundedness with respect to the switching action is given. Nevertheless, in order to help the driver, a haptic warning is switched on, after the automatic driving had been deactivated. The warning consists in steering suggestions. These suggestions are generated as asymmetric vibrations of the steering wheel. They consist in a DC motor torque, following a modified sinus signal which has a non-zero mean value. This signal indicates which way the driver should turn the steering wheel in order to return to the centre of the lane. However, these vibrations are a high frequency torque in open loop and are not able to keep the vehicle on the lane without the driver's action on the steering wheel.[1] N.Minoiu, M. Netto and S. Mammar, “A switched optimized approach for road departure avoidance'”, Proc. IEEE Conf. Intelligent Transportation Systems, 2006, pp. 133-138.

An evaluation of the robustness of existing supervised machine learning approaches to the classification of emotions in speech

In this study, the robustness of approaches to the automatic classification of emotions in speech is addressed. Among the many types of emotions that exist, two groups of emotions are considered, adult-to-adult acted vocal expressions of common types of emotions like happiness, sadness, and anger and adult-to-infant vocal expressions of affective intents also known as “motherese”. Specifically, we estimate the generalization capability of two feature extraction approaches, the approach developed for Sony’s robotic dog AIBO (AIBO) and the segment-based approach (SBA) of [Shami, M., Kamel, M., 2005. Segment-based approach to the recognition of emotions in speech. In: IEEE Conf. on Multimedia and Expo (ICME05), Amsterdam, The Netherlands]. Three machine learning approaches are considered, K-nearest neighbors (KNN), Support vector machines (SVM) and Ada-boosted decision trees and four emotional speech databases are employed, Kismet, BabyEars, Danish, and Berlin databases. Single corpus experiments show that the considered feature extraction approaches AIBO and SBA are competitive on the four databases considered and that their performance is comparable with previously published results on the same databases. The best choice of machine learning algorithm seems to depend on the feature extraction approach considered. Multi-corpus experiments are performed with the Kismet–BabyEars and the Danish–Berlin database pairs that contain parallel emotional classes. Automatic clustering of the emotional classes in the database pairs shows that the patterns behind the emotions in the Kismet–BabyEars pair are less database dependent than the patterns in the Danish–Berlin pair. In off-corpus testing the classifier is trained on one database of a pair and tested on the other. This provides little improvement over baseline classification. In integrated corpus testing, however, the classifier is machine learned on the merged databases and this gives promisingly robust classification results, which suggest that emotional corpora with parallel emotion classes recorded under different conditions can be used to construct a single classifier capable of distinguishing the emotions in the merged corpora. Such a classifier is more robust than a classifier learned on a single corpus as it can recognize more varied expressions of the same emotional classes. These findings suggest that the existing approaches for the classification of emotions in speech are efficient enough to handle larger amounts of training data without any reduction in classification accuracy.