Hidden Conditional Random Fields for Visual Speech Recognition

In this paper we present the application of Hidden Conditional Random Fields (HCRFs) to modelling speech for visual speech recognition. HCRFs may be easily adapted to model long range dependencies across an observation sequence. As a result visual word recognition performance can be improved as the model is able to take more of a contextual approach to generating state sequences. Results are presented from a speaker-dependent, isolated digit, visual speech recognition task using comparisons with a baseline HMM system. We firstly illustrate that word recognition rates on clean video using HCRFs can be improved by increasing the number of past and future observations being taken into account by each state. Secondly we compare model performances using various levels of video compression on the test set. As far as we are aware this is the first attempted use of HCRFs for visual speech recognition.

Implicit and explicit processes in a hypothesis testing task

We present the results of two experiments investigating the factors that determine responding on the pseudo-diagnosticity task. In Expt I we manipulated people's beliefs about the degree to which an initial piece of evidence supported a focal hypothesis and found decreased pseudo-diagnostic (PD) responding when the evidence offered low support for the focal hypothesis. In Expt 2 we manipulated the instructions given to participants. We found that instructions to select evidence to help decide between the focal and the complementary hypotheses produced fewer PD responses than both instructions to decide whether the focal hypothesis was the case and instructions to decide whether its complement was the case. The results are interpreted within the framework of recent dual process theories of reasoning.

Systematic random sampling of the comet assay

The comet assay is a technique used to quantify DNA damage and repair at a cellular level. In the assay, cells are embedded in agarose and the cellular content is stripped away leaving only the DNA trapped in an agarose cavity which can then be electrophoresed. The damaged DNA can enter the agarose and migrate while the undamaged DNA cannot and is retained. DNA damage is measured as the proportion of the migratory ‘tail’ DNA compared to the total DNA in the cell. The fundamental basis of these arbitrary values is obtained in the comet acquisition phase using fluorescence microscopy with a stoichiometric stain in tandem with image analysis software. Current methods deployed in such an acquisition are expected to be both objectively and randomly obtained. In this paper we examine the ‘randomness’ of the acquisition phase and suggest an alternative method that offers both objective and unbiased comet selection. In order to achieve this, we have adopted a survey sampling approach widely used in stereology, which offers a method of systematic random sampling (SRS). This is desirable as it offers an impartial and reproducible method of comet analysis that can be used both manually or automated. By making use of an unbiased sampling frame and using microscope verniers, we are able to increase the precision of estimates of DNA damage. Results obtained from a multiple-user pooled variation experiment showed that the SRS technique attained a lower variability than that of the traditional approach. The analysis of a single user with repetition experiment showed greater individual variances while not being detrimental to overall averages. This would suggest that the SRS method offers a better reflection of DNA damage for a given slide and also offers better user reproducibility.

Validity of a 1 minute walk test for children with cerebral palsy.

The concurrent validity of a 1 minute walk test at a child's maximum walking speed was assessed in children with bilateral spastic cerebral palsy (BSCP). The distance covered during the 1 minute walk test was compared with the children's gross motor function as assessed by the Gross Motor Function Measure (GMFM). Twenty-four male and 10 female children with CP (mean age 11y, range 4 to 16y) participated in the study. Gross Motor Function Classification System (GMFCS) levels were; level I (n=3), level II (n=17), level III (n=10), and level IV (n=4). Participants had clinical diagnoses of symmetrical diplegia (n=19), asymmetrical diplegia (n=14), and quadriplegia (n=1). Results showed a significant correlation between GMFM score and the distance covered during the 1 minute walk (r=0.92; p<0.001). There was also a significant decrease in the distance walked with increasing GMFCS level (p<0.001). We concluded that the 1 minute walk test is a valid measure for assessing functional ability in children with ambulatory BSCP. Its cost-effectiveness and user friendliness make it a potentially useful tool in the clinical setting. Further study needs to address its reliability and ability to detect change over time.

Structural change in multipartite entanglement sharing: A random matrix approach

We study the typical entanglement properties of a system comprising two independent qubit environments interacting via a shuttling ancilla. The initial preparation of the environments is modeled using random matrix techniques. The entanglement measure used in our study is then averaged over many histories of randomly prepared environmental states. Under a Heisenberg interaction model, the average entanglement between the ancilla and one of the environments remains constant, regardless of the preparation of the latter and the details of the interaction. We also show that, upon suitable kinematic and dynamical changes in the ancillaenvironment subsystems, the entanglement-sharing structure undergoes abrupt modifications associated with a change in the multipartite entanglement class of the overall system's state. These results are invariant with respect to the randomized initial state of the environments.

Infections and vaccinations as risk factors for childhood type I (insulin-dependent) diabetes mellitus: a multicentre case-control investigation.

AIMS/HYPOTHESIS: To determine if vaccinations and infections are associated with the subsequent risk of Type I (insulin-dependent) diabetes mellitus in childhood. METHOD: Seven centres in Europe with access to population-based registers of children with Type I diabetes diagnosed under 15 years of age participated in a case-control study of environmental risk factors. Control children were chosen at random in each centre either from population registers or from schools and policlinics. Data on maternal and neonatal infections, common childhood infections and vaccinations were obtained for 900 cases and 2302 control children from hospital and clinic records and from parental responses to a questionnaire or interview. RESULTS: Infections early in the child's life noted in the hospital record were found to be associated with an increased risk of diabetes, although the odds ratio of 1.61 (95% confidence limits 1.11, 2.33) was significant only after adjustment for confounding variables. None of the common childhood infectious diseases was found to be associated with diabetes and neither was there evidence that any common childhood vaccination modified the risk of diabetes. Pre-school day-care attendance, a proxy measure for total infectious disease exposure in early childhood, was found, however, to be inversely associated with diabetes, with a pooled odds ratio of 0.59 (95% confidence limits 0.46, 0.76) after adjustment for confounding variables. CONCLUSION/INTERPRETATION: It seems likely that the explanation for these contrasting findings of an increased risk associated with perinatal infections coupled with a protective effect of pre-school day care lies in the age-dependent modifying influence of infections on the developing immune system.