Feature Selection for Pose Invariant Lip Biometrics

For the first time in this paper the authors present results showing the effect of out of plane speaker head pose variation on a lip biometric based speaker verification system. Using appearance DCT based features, they adopt a Mutual Information analysis technique to highlight the class discriminant DCT components most robust to changes in out of plane pose. Experiments are conducted using the initial phase of a new multi view Audio-Visual database designed for research and development of pose-invariant speech and speaker recognition. They show that verification performance can be improved by substituting higher order horizontal DCT components for vertical, particularly in the case of a train/test pose angle mismatch.

Modelling the mechanical response of the Reed-mouthpiece-lip system of a clarinet. Part I. A one-dimensional distributed model

The motion of a clarinet reed that is clamped to a mouthpiece and supported by a lip is simulated in the time-domain using finite difference methods. The reed is modelled as a bar with non-uniform cross section, and is described using a one-dimensional, fourth-order partial differential equation. The interactions with the mouthpiece Jay and the player's lip are taken into account by incorporating conditional contact forces in the bar equation. The model is completed by clamped-free boundary conditions for the reed. An implicit finite difference method is used for discretising the system, and values for the physical parameters are chosen both from laboratory measurements and by accurate tuning of the numerical simulations. The accuracy of the numerical system is assessed through analysis of frequency warping effects and of resonance estimation. Finally, the mechanical properties of the system are studied by analysing its response to external driving forces. In particular, the effects of reed curling are investigated.

Modelling the mechanical response of the reed-mouthpiece-lip system of a clarinet. Part II: A lumped model approximation

A non-linear lumped model of the reed-mouthpiece-lip system of a clarinet is formulated, in which the lumped parameters are derived from numerical experiments with a finite-difference simulation based on a distributed reed model. The effective stiffness per unit area is formulated as a function of the pressure signal driving the reed, in order to simulate the effects of the reed bending against the lay, and mass and damping terms are added as a first approximation to the dynamic behaviour of the reed. A discrete-time formulation is presented, and its response is compared to that of the distributed model. In addition, the lumped model is applied in the simulation of clarinet tones, enabling the analysis of the effects of using a pressure-dependent stiffness per unit area on sustained oscillations. The analysed effects and features are in qualitative agreement with players' experiences and experimental results obtained in prior studies.

Gender Classification via Lips: Static and Dynamic Features

Automatic gender classification has many security and commercial applications. Various modalities have been investigated for gender classification with face-based classification being the most popular. In some real-world scenarios the face may be partially occluded. In these circumstances a classification based on individual parts of the face known as local features must be adopted. We investigate gender classification using lip movements. We show for the first time that important gender specific information can be obtained from the way in which a person moves their lips during speech. Furthermore our study indicates that the lip dynamics during speech provide greater gender discriminative information than simply lip appearance. We also show that the lip dynamics and appearance contain complementary gender information such that a model which captures both traits gives the highest overall classification result. We use Discrete Cosine Transform based features and Gaussian Mixture Modelling to model lip appearance and dynamics and employ the XM2VTS database for our experiments. Our experiments show that a model which captures lip dynamics along with appearance can improve gender classification rates by between 16-21% compared to models of only lip appearance.

How Does Justice Sound? Enunciative Modalities, Law’s Logos and Woman’s Voice

In the throes of her mimetic exposure of the lie of phallocratic discursive unity in 'Speculum of the Other Woman', Irigaray paused on the impossibility of woman’s voice and remarked that ‘it [was] still better to speak only in riddles, allusions, hints, parables.’ Even if asked to clarify a few points. Even if people plead that they just don’t understand. After all, she said, ‘they never have understood.’ (Irigaray 1985, 143).

That the law has never understood a uniquely feminine narrative is hardly controversial, but that this erasure continues to have real and substantive consequences for justice is a reality that feminists have been compelled to remain vigilant in exposing. How does the authority of the word compound law’s exclusionary matrix? How does law remain impervious to woman’s voice and how might it hear woman’s voice? Is there capacity for a dialogic engagement between woman, parler femme, and law?

This paper will explore these questions with particular reference to the experience of women testifying to trauma during the rape trial. It will argue that a logically linked historical genealogy can be traced through which law has come to posit itself as an originary discourse by which thinking is very much conflated with being, or in other terms, law is conflated with justice. This has consequences both for women’s capacity to speak or represent the harm of rape to law, but also for law’s ability to ‘hear’ woman’s voice and objectively adjudicate in cases of rape. It will suggest that justice requires law acknowledge the presence of two distinct and different subjects and that this must be done not only at the symbolic level but also at the level of the parole, syntax and discourse.

Decoding semantics across fMRI sessions with different stimulus modalities:a practical MVPA study

Both embodied and symbolic accounts of conceptual organization would predict partial sharing and partial differentiation between the neural activations seen for concepts activated via different stimulus modalities. But cross-participant and cross-session variability in BOLD activity patterns makes analyses of such patterns with MVPA methods challenging. Here, we examine the effect of cross-modal and individual variation on the machine learning analysis of fMRI data recorded during a word property generation task. We present the same set of living and non-living concepts (land-mammals, or work tools) to a cohort of Japanese participants in two sessions: the first using auditory presentation of spoken words; the second using visual presentation of words written in Japanese characters. Classification accuracies confirmed that these semantic categories could be detected in single trials, with within-session predictive accuracies of 80-90%. However cross-session prediction (learning from auditory-task data to classify data from the written-word-task, or vice versa) suffered from a performance penalty, achieving 65-75% (still individually significant at p « 0.05). We carried out several follow-on analyses to investigate the reason for this shortfall, concluding that distributional differences in neither time nor space alone could account for it. Rather, combined spatio-temporal patterns of activity need to be identified for successful cross-session learning, and this suggests that feature selection strategies could be modified to take advantage of this.

The lip:A histologic and analytical approach of relevance to esthetic plastic surgery

This is a study done for the first time to understand the histologic features of the lip deeply, quantitatively, and analytically and identify the differences between the outer, intermediate, and inner parts of the lips. Previous researchers did not tackle the histologic features of the lip from this point of view. Half lip was obtained from different well-preserved cadavers where the upper lip, lower lip, and lip angle were evaluated by coronal histologic sections. A total of 43 slides were studied and photographed using light and digital microscopes (Nikon COOLSCOPE, Nikon Corporation, Tokyo, Japan), respectively. The cadavers (26 men and 17 women) were in the age group of 45 to 65 years old, and older than 65 years. Data were entered on a study pro forma and statistically analyzed. Normal histologic features of the upper lip, intermediate, and the inner lip sections were observed. Fibrous septations that sometimes had muscular components inserting into them could be identified. In between these septations, there were loose areas (chambers). The mean numbers of chambers identified in the upper and lower lips were higher in the red areas, and more septations and chambers were identified in the lower lip. New histologic observations were classified into types 1 to 7. Significant relations were identified between the total number of septations and chambers in relation to age group and sex. Higher means of septations and chambers were detected in the age group older than 65 years, and in women as a whole. In conclusion, the findings detected in this work could explain the congenital lip pits that are familial or syndromic and many post-lip augmentation complications. © 2009 Lippincott Williams & Wilkins, Inc.

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.

Combinations of Prompt and Target Response Modalities and Their Impact on Emergent Relations and Teaching Efficiency.

In the present study, native Spanish speakers were taught a small English vocabulary (Spanish-to-English intraverbals). Four different training conditions were created by combining textual and echoic prompts with written and vocal target responses. The efficiency of each training condition was examined by analysing emergent relations (i.e., tacts) and the total number of sessions required to reach mastery under each training condition. All combinations of prompt-response modalities generated increases in correct responding on tests for emergent relations but when target responses were written, mastery criterion was reached faster. Results are discussed in terms of efficiency for emergent relations and recommendations for future directions are provided.

Using a Discrete Hidden Markov Model Kernel for lip-based biometric identification

In this paper, a novel and effective lip-based biometric identification approach with the Discrete Hidden Markov Model Kernel (DHMMK) is developed. Lips are described by shape features (both geometrical and sequential) on two different grid layouts: rectangular and polar. These features are then specifically modeled by a DHMMK, and learnt by a support vector machine classifier. Our experiments are carried out in a ten-fold cross validation fashion on three different datasets, GPDS-ULPGC Face Dataset, PIE Face Dataset and RaFD Face Dataset. Results show that our approach has achieved an average classification accuracy of 99.8%, 97.13%, and 98.10%, using only two training images per class, on these three datasets, respectively. Our comparative studies further show that the DHMMK achieved a 53% improvement against the baseline HMM approach. The comparative ROC curves also confirm the efficacy of the proposed lip contour based biometrics learned by DHMMK. We also show that the performance of linear and RBF SVM is comparable under the frame work of DHMMK.

Investigation into DCT Feature Selection for Visual Lip-Based Biometric Authentication

This paper investigated using lip movements as a behavioural biometric for person authentication. The system was trained, evaluated and tested using the XM2VTS dataset, following the Lausanne Protocol configuration II. Features were selected from the DCT coefficients of the greyscale lip image. This paper investigated the number of DCT coefficients selected, the selection process, and static and dynamic feature combinations. Using a Gaussian Mixture Model - Universal Background Model framework an Equal Error Rate of 2.20% was achieved during evaluation and on an unseen test set a False Acceptance Rate of 1.7% and False Rejection Rate of 3.0% was achieved. This compares favourably with face authentication results on the same dataset whilst not being susceptible to spoofing attacks.