Exploiting Phonological Constraints for Handshape Inference in ASL Video

Handshape is a key articulatory parameter in sign language, and thus handshape recognition from signing video is essential for sign recognition and retrieval. Handshape transitions within monomorphemic lexical signs (the largest class of signs in signed languages) are governed by phonological rules. For example, such transitions normally involve either closing or opening of the hand (i.e., to exclusively use either folding or unfolding of the palm and one or more fingers). Furthermore, akin to allophonic variations in spoken languages, both inter- and intra- signer variations in the production of specific handshapes are observed. We propose a Bayesian network formulation to exploit handshape co-occurrence constraints, also utilizing information about allophonic variations to aid in handshape recognition. We propose a fast non-rigid image alignment method to gain improved robustness to handshape appearance variations during computation of observation likelihoods in the Bayesian network. We evaluate our handshape recognition approach on a large dataset of monomorphemic lexical signs. We demonstrate that leveraging linguistic constraints on handshapes results in improved handshape recognition accuracy. As part of the overall project, we are collecting and preparing for dissemination a large corpus (three thousand signs from three native signers) of American Sign Language (ASL) video. The video have been annotated using SignStream® [Neidle et al.] with labels for linguistic information such as glosses, morphological properties and variations, and start/end handshapes associated with each ASL sign.

SymbolDesign: A User-centered Method to Design Pen-based Interfaces and Extend the Functionality of Pointer Input Devices

A method called "SymbolDesign" is proposed that can be used to design user-centered interfaces for pen-based input devices. It can also extend the functionality of pointer input devices such as the traditional computer mouse or the Camera Mouse, a camera-based computer interface. Users can create their own interfaces by choosing single-stroke movement patterns that are convenient to draw with the selected input device and by mapping them to a desired set of commands. A pattern could be the trace of a moving finger detected with the Camera Mouse or a symbol drawn with an optical pen. The core of the SymbolDesign system is a dynamically created classifier, in the current implementation an artificial neural network. The architecture of the neural network automatically adjusts according to the complexity of the classification task. In experiments, subjects used the SymbolDesign method to design and test the interfaces they created, for example, to browse the web. The experiments demonstrated good recognition accuracy and responsiveness of the user interfaces. The method provided an easily-designed and easily-used computer input mechanism for people without physical limitations, and, with some modifications, has the potential to become a computer access tool for people with severe paralysis.

Spatiotemporal Gesture Segmentation

Spotting patterns of interest in an input signal is a very useful task in many different fields including medicine, bioinformatics, economics, speech recognition and computer vision. Example instances of this problem include spotting an object of interest in an image (e.g., a tumor), a pattern of interest in a time-varying signal (e.g., audio analysis), or an object of interest moving in a specific way (e.g., a human's body gesture). Traditional spotting methods, which are based on Dynamic Time Warping or hidden Markov models, use some variant of dynamic programming to register the pattern and the input while accounting for temporal variation between them. At the same time, those methods often suffer from several shortcomings: they may give meaningless solutions when input observations are unreliable or ambiguous, they require a high complexity search across the whole input signal, and they may give incorrect solutions if some patterns appear as smaller parts within other patterns. In this thesis, we develop a framework that addresses these three problems, and evaluate the framework's performance in spotting and recognizing hand gestures in video. The first contribution is a spatiotemporal matching algorithm that extends the dynamic programming formulation to accommodate multiple candidate hand detections in every video frame. The algorithm finds the best alignment between the gesture model and the input, and simultaneously locates the best candidate hand detection in every frame. This allows for a gesture to be recognized even when the hand location is highly ambiguous. The second contribution is a pruning method that uses model-specific classifiers to reject dynamic programming hypotheses with a poor match between the input and model. Pruning improves the efficiency of the spatiotemporal matching algorithm, and in some cases may improve the recognition accuracy. The pruning classifiers are learned from training data, and cross-validation is used to reduce the chance of overpruning. The third contribution is a subgesture reasoning process that models the fact that some gesture models can falsely match parts of other, longer gestures. By integrating subgesture reasoning the spotting algorithm can avoid the premature detection of a subgesture when the longer gesture is actually being performed. Subgesture relations between pairs of gestures are automatically learned from training data. The performance of the approach is evaluated on two challenging video datasets: hand-signed digits gestured by users wearing short sleeved shirts, in front of a cluttered background, and American Sign Language (ASL) utterances gestured by ASL native signers. The experiments demonstrate that the proposed method is more accurate and efficient than competing approaches. The proposed approach can be generally applied to alignment or search problems with multiple input observations, that use dynamic programming to find a solution.

Adaptive Bacteria Colony Picking in Unstructured Environments Using Intensity Histogram and Unascertained LS-SVM Classifier

Features analysis is an important task which can significantly affect the performance of automatic bacteria colony picking. Unstructured environments also affect the automatic colony screening. This paper presents a novel approach for adaptive colony segmentation in unstructured environments by treating the detected peaks of intensity histograms as a morphological feature of images. In order to avoid disturbing peaks, an entropy based mean shift filter is introduced to smooth images as a preprocessing step. The relevance and importance of these features can be determined in an improved support vector machine classifier using unascertained least square estimation. Experimental results show that the proposed unascertained least square support vector machine (ULSSVM) has better recognition accuracy than the other state-of-the-art techniques, and its training process takes less time than most of the traditional approaches presented in this paper.

Robust bimodal person identification using face and speech with limited training data and corruption of both modalities

This paper presents a novel method of audio-visual fusion for person identification where both the speech and facial modalities may be corrupted, and there is a lack of prior knowledge about the corruption. Furthermore, we assume there is a limited amount of training data for each modality (e.g., a short training speech segment and a single training facial image for each person). A new representation and a modified cosine similarity are introduced for combining and comparing bimodal features with limited training data as well as vastly differing data rates and feature sizes. Optimal feature selection and multicondition training are used to reduce the mismatch between training and testing, thereby making the system robust to unknown bimodal corruption. Experiments have been carried out on a bimodal data set created from the SPIDRE and AR databases with variable noise corruption of speech and occlusion in the face images. The new method has demonstrated improved recognition accuracy.

Gait period estimation algorithm based on the angles of extremities

Gait period estimation is an important step in the gait recognition framework. In this paper, we propose a new gait cycle detection method based on the angles of extreme points of both legs. In addition to that, to further improve the estimation of the gait period, the proposed algorithm divides the gait sequence into sections before identifying the maximum values. The proposed algorithm is scale invariant and less dependent on the silhouette shape. The performance of the proposed method was evaluated using the OU-ISIR speed variation gait database. The experimental results show that the proposed method achieved 90.2% gait recognition accuracy and outperforms previous methods found in the literature with the second best only achieved 67.65% accuracy.

The generation effect and word learning: a test of the effect of the language experience approach versus the text approach in the acquisition of new reading vocabulary

A class of twenty-two grade one children was tested to determine their reading levels using the Stanford Diagnostic Reading Achievement Test. Based on these results and teacher input the students were paired according to reading ability. The students ages ranged from six years four months to seven years four months at the commencement of the study. Eleven children were assigned to the language experience group and their partners became the text group. Each member of the language experience group generated a list of eight to be learned words. The treatment consisted of exposing the student to a given word three times per session for ten sessions, over a period of five days. The dependent variables consisted of word identification speed, word identification accuracy, and word recognition accuracy. Each member of the text group followed the same procedure using his/her partner's list of words. Upon completion of this training, the entire process was repeated with members of the text group from the first part becoming members of the language experience group and vice versa. The results suggest that generally speaking language experience words are identified faster than text words but that there is no difference in the rate at which these words are learned. Language experience words may be identified faster because the auditory-semantic information is more readily available in them than in text words. The rate of learning in both types of words, however, may be dictated by the orthography of the to be learned word.

An Investigation of the Role of Attention in the Cross-Race Effect: An Ecological Approach

The current set of studies was conducted to examine the cross-race effect (CRE), a phenomenon commonly found in the face perception literature. The CRE is evident when participants display better own-race face recognition accuracy than other-race recognition accuracy (e.g. Ackerman et al., 2006). Typically the cross-race effect is attributed to perceptual expertise, (i.e., other-race faces are processed less holistically; Michel, Rossion, Han, Chung & Caldara, 2006), and the social cognitive model (i.e., other-race faces are processed at the categorical level by virtue of being an out-group member; Hugenberg, Young, Bernstein, & Sacco, 2010). These effects may be mediated by differential attention. I investigated whether other-race faces are disregarded and, consequently, not remembered as accurately as own-race (in-group) faces. In Experiment 1, I examined how the magnitude of the CRE differed when participants learned individual faces sequentially versus when they learned multiple faces simultaneously in arrays comprising faces and objects. I also examined how the CRE differed when participants recognized individual faces presented sequentially versus in arrays of eight faces. Participants’ recognition accuracy was better for own-race faces than other-race faces regardless of familiarization method. However, the difference between own- and other-race accuracy was larger when faces were familiarized sequentially in comparison to familiarization with arrays. Participants’ response patterns during testing differed depending on the combination of familiarization and testing method. Participants had more false alarms for other-race faces than own-race faces if they learned faces sequentially (regardless of testing strategy); if participants learned faces in arrays, they had more false alarms for other-race faces than own-races faces if ii i they were tested with sequentially presented faces. These results are consistent with the perceptual expertise model in that participants were better able to use the full two seconds in the sequential task for own-race faces, but not for other-race faces. The purpose of Experiment 2 was to examine participants’ attentional allocation in complex scenes. Participants were shown scenes comprising people in real places, but the head stimuli used in Experiment 1 were superimposed onto the bodies in each scene. Using a Tobii eyetracker, participants’ looking time for both own- and other-race faces was evaluated to determine whether participants looked longer at own-race faces and whether individual differences in looking time correlated with individual differences in recognition accuracy. The results of this experiment demonstrated that although own-race faces were preferentially attended to in comparison to other-race faces, individual differences in looking time biases towards own-race faces did not correlate with individual differences in own-race recognition advantages. These results are also consistent with perceptual expertise, as it seems that the role of attentional biases towards own-race faces is independent of the cognitive processing that occurs for own-race faces. All together, these results have implications for face perception tasks that are performed in the lab, how accurate people may be when remembering faces in the real world, and the accuracy and patterns of errors in eyewitness testimony.

L'effort associé à la reconnaissance de la parole chez les adultes et les personnes aînées

L’objectif principal de cette thèse était de quantifier et comparer l’effort requis pour reconnaître la parole dans le bruit chez les jeunes adultes et les personnes aînées ayant une audition normale et une acuité visuelle normale (avec ou sans lentille de correction de la vue). L’effort associé à la perception de la parole est lié aux ressources attentionnelles et cognitives requises pour comprendre la parole. La première étude (Expérience 1) avait pour but d’évaluer l’effort associé à la reconnaissance auditive de la parole (entendre un locuteur), tandis que la deuxième étude (Expérience 2) avait comme but d’évaluer l’effort associé à la reconnaissance auditivo-visuelle de la parole (entendre et voir le visage d’un locuteur). L’effort fut mesuré de deux façons différentes. D’abord par une approche comportementale faisant appel à un paradigme expérimental nommé double tâche. Il s’agissait d’une tâche de reconnaissance de mot jumelée à une tâche de reconnaissance de patrons vibro-tactiles. De plus, l’effort fut quantifié à l’aide d’un questionnaire demandant aux participants de coter l’effort associé aux tâches comportementales. Les deux mesures d’effort furent utilisées dans deux conditions expérimentales différentes : 1) niveau équivalent – c'est-à-dire lorsque le niveau du bruit masquant la parole était le même pour tous les participants et, 2) performance équivalente – c'est-à-dire lorsque le niveau du bruit fut ajusté afin que les performances à la tâche de reconnaissance de mots soient identiques pour les deux groupes de participant. Les niveaux de performance obtenus pour la tâche vibro-tactile ont révélé que les personnes aînées fournissent plus d’effort que les jeunes adultes pour les deux conditions expérimentales, et ce, quelle que soit la modalité perceptuelle dans laquelle les stimuli de la parole sont présentés (c.-à.-d., auditive seulement ou auditivo-visuelle). Globalement, le ‘coût’ associé aux performances de la tâche vibro-tactile était au plus élevé pour les personnes aînées lorsque la parole était présentée en modalité auditivo-visuelle. Alors que les indices visuels peuvent améliorer la reconnaissance auditivo-visuelle de la parole, nos résultats suggèrent qu’ils peuvent aussi créer une charge additionnelle sur les ressources utilisées pour traiter l’information. Cette charge additionnelle a des conséquences néfastes sur les performances aux tâches de reconnaissance de mots et de patrons vibro-tactiles lorsque celles-ci sont effectuées sous des conditions de double tâche. Conformément aux études antérieures, les coefficients de corrélations effectuées à partir des données de l’Expérience 1 et de l’Expérience 2 soutiennent la notion que les mesures comportementales de double tâche et les réponses aux questionnaires évaluent différentes dimensions de l’effort associé à la reconnaissance de la parole. Comme l’effort associé à la perception de la parole repose sur des facteurs auditifs et cognitifs, une troisième étude fut complétée afin d’explorer si la mémoire auditive de travail contribue à expliquer la variance dans les données portant sur l’effort associé à la perception de la parole. De plus, ces analyses ont permis de comparer les patrons de réponses obtenues pour ces deux facteurs après des jeunes adultes et des personnes aînées. Pour les jeunes adultes, les résultats d’une analyse de régression séquentielle ont démontré qu’une mesure de la capacité auditive (taille de l’empan) était reliée à l’effort, tandis qu’une mesure du traitement auditif (rappel alphabétique) était reliée à la précision avec laquelle les mots étaient reconnus lorsqu’ils étaient présentés sous les conditions de double tâche. Cependant, ces mêmes relations n’étaient pas présentes dans les données obtenues pour le groupe de personnes aînées ni dans les données obtenues lorsque les tâches de reconnaissance de la parole étaient effectuées en modalité auditivo-visuelle. D’autres études sont nécessaires pour identifier les facteurs cognitifs qui sous-tendent l’effort associé à la perception de la parole, et ce, particulièrement chez les personnes aînées.

Apprentissage d'espaces sémantiques

Dans cette dissertation, nous présentons plusieurs techniques d’apprentissage d’espaces sémantiques pour plusieurs domaines, par exemple des mots et des images, mais aussi à l’intersection de différents domaines. Un espace de représentation est appelé sémantique si des entités jugées similaires par un être humain, ont leur similarité préservée dans cet espace. La première publication présente un enchaînement de méthodes d’apprentissage incluant plusieurs techniques d’apprentissage non supervisé qui nous a permis de remporter la compétition “Unsupervised and Transfer Learning Challenge” en 2011. Le deuxième article présente une manière d’extraire de l’information à partir d’un contexte structuré (177 détecteurs d’objets à différentes positions et échelles). On montrera que l’utilisation de la structure des données combinée à un apprentissage non supervisé permet de réduire la dimensionnalité de 97% tout en améliorant les performances de reconnaissance de scènes de +5% à +11% selon l’ensemble de données. Dans le troisième travail, on s’intéresse à la structure apprise par les réseaux de neurones profonds utilisés dans les deux précédentes publications. Plusieurs hypothèses sont présentées et testées expérimentalement montrant que l’espace appris a de meilleures propriétés de mixage (facilitant l’exploration de différentes classes durant le processus d’échantillonnage). Pour la quatrième publication, on s’intéresse à résoudre un problème d’analyse syntaxique et sémantique avec des réseaux de neurones récurrents appris sur des fenêtres de contexte de mots. Dans notre cinquième travail, nous proposons une façon d’effectuer de la recherche d’image ”augmentée” en apprenant un espace sémantique joint où une recherche d’image contenant un objet retournerait aussi des images des parties de l’objet, par exemple une recherche retournant des images de ”voiture” retournerait aussi des images de ”pare-brises”, ”coffres”, ”roues” en plus des images initiales.

Feature Extraction Methods Based on Linear Predictive Coding and Wavelet Packet Decomposition for Recognizing Spoken Words in Malayalam

Speech signals are one of the most important means of communication among the human beings. In this paper, a comparative study of two feature extraction techniques are carried out for recognizing speaker independent spoken isolated words. First one is a hybrid approach with Linear Predictive Coding (LPC) and Artificial Neural Networks (ANN) and the second method uses a combination of Wavelet Packet Decomposition (WPD) and Artificial Neural Networks. Voice signals are sampled directly from the microphone and then they are processed using these two techniques for extracting the features. Words from Malayalam, one of the four major Dravidian languages of southern India are chosen for recognition. Training, testing and pattern recognition are performed using Artificial Neural Networks. Back propagation method is used to train the ANN. The proposed method is implemented for 50 speakers uttering 20 isolated words each. Both the methods produce good recognition accuracy. But Wavelet Packet Decomposition is found to be more suitable for recognizing speech because of its multi-resolution characteristics and efficient time frequency localizations

Automated Transcription System for Malayalam Language

Malayalam is one of the 22 scheduled languages in India with more than 130 million speakers. This paper presents a report on the development of a speaker independent, continuous transcription system for Malayalam. The system employs Hidden Markov Model (HMM) for acoustic modeling and Mel Frequency Cepstral Coefficient (MFCC) for feature extraction. It is trained with 21 male and female speakers in the age group ranging from 20 to 40 years. The system obtained a word recognition accuracy of 87.4% and a sentence recognition accuracy of 84%, when tested with a set of continuous speech data.

Robust multimodal face and fingerprint fusion in the presence of spoofing attacks

Anti-spoofing is attracting growing interest in biometrics, considering the variety of fake materials and new means to attack biometric recognition systems. New unseen materials continuously challenge state-of-the-art spoofing detectors, suggesting for additional systematic approaches to target anti-spoofing. By incorporating liveness scores into the biometric fusion process, recognition accuracy can be enhanced, but traditional sum-rule based fusion algorithms are known to be highly sensitive to single spoofed instances. This paper investigates 1-median filtering as a spoofing-resistant generalised alternative to the sum-rule targeting the problem of partial multibiometric spoofing where m out of n biometric sources to be combined are attacked. Augmenting previous work, this paper investigates the dynamic detection and rejection of livenessrecognition pair outliers for spoofed samples in true multi-modal configuration with its inherent challenge of normalisation. As a further contribution, bootstrap aggregating (bagging) classifiers for fingerprint spoof-detection algorithm is presented. Experiments on the latest face video databases (Idiap Replay- Attack Database and CASIA Face Anti-Spoofing Database), and fingerprint spoofing database (Fingerprint Liveness Detection Competition 2013) illustrate the efficiency of proposed techniques.

Visual perception of movement kinematics and the acquisition of "action prototypes"

Recognizing a class of movements as belonging to a "nominal" action category, such as walking, running, or throwing, is a fundamental human ability. Three experiments were undertaken to test the hypothesis that common ("prototypical") features of moving displays could be learned by observation. Participants viewed moving stick-figure displays resembling forearm flexion movements in the saggital plane. Four displays (presentation displays) were first presented in which one or more movement dimensions were combined with 2 respective cues: direction (up, down), speed (fast, slow), and extent (long, short). Eight test displays were then shown, and the observer indicated whether each test display was like or unlike those previously seen. The results showed that without corrective feedback, a single cue (e.g., up or down) could be correctly recognized, on average, with the proportion correct between .66 and .87. When two cues were manipulated (e.g., up and slow), recognition accuracy remained high, ranging between .72 and .89. Three-cue displays were also easily identified. These results provide the first empirical demonstration of action-prototype learning for categories of human action and show how apparently complex kinematic patterns can be categorized in terms of common features or cues. It was also shown that probability of correct recognition of kinematic properties was reduced when the set of 4 presentation displays were more variable with respect to their shared kinematic property, such as speed or amplitude. Finally, while not conclusive, the results (from 2 of the 3 experiments) did suggest that similarity (or "likeness") with respect to a common kinematic property (or properties) is more easily recognized than dissimilarity.

Learning and detecting activities from movement trajectories using the hierarchical hidden Markov model

Directly modeling the inherent hierarchy and shared structures of human behaviors, we present an application of the hierarchical hidden Markov model (HHMM) for the problem of activity recognition. We argue that to robustly model and recognize complex human activities, it is crucial to exploit both the natural hierarchical decomposition and shared semantics embedded in the movement trajectories. To this end, we propose the use of the HHMM, a rich stochastic model that has been recently extended to handle shared structures, for representing and recognizing a set of complex indoor activities. Furthermore, in the need of real-time recognition, we propose a Rao-Blackwellised particle filter (RBPF) that efficiently computes the filtering distribution at a constant time complexity for each new observation arrival. The main contributions of this paper lie in the application of the shared-structure HHMM, the estimation of the model's parameters at all levels simultaneously, and a construction of an RBPF approximate inference scheme. The experimental results in a real-world environment have confirmed our belief that directly modeling shared structures not only reduces computational cost, but also improves recognition accuracy when compared with the tree HHMM and the flat HMM.