Speech Recognition with unknown partial feature corruption - a review of the union model

This paper provides a summary of our studies on robust speech recognition based on a new statistical approach – the probabilistic union model. We consider speech recognition given that part of the acoustic features may be corrupted by noise. The union model is a method for basing the recognition on the clean part of the features, thereby reducing the effect of the noise on recognition. To this end, the union model is similar to the missing feature method. However, the two methods achieve this end through different routes. The missing feature method usually requires the identity of the noisy data for noise removal, while the union model combines the local features based on the union of random events, to reduce the dependence of the model on information about the noise. We previously investigated the applications of the union model to speech recognition involving unknown partial corruption in frequency band, in time duration, and in feature streams. Additionally, a combination of the union model with conventional noise-reduction techniques was studied, as a means of dealing with a mixture of known or trainable noise and unknown unexpected noise. In this paper, a unified review, in the context of dealing with unknown partial feature corruption, is provided into each of these applications, giving the appropriate theory and implementation algorithms, along with an experimental evaluation.

COGNITIVE ASSESSMENT USING THE KINARM EXOSKELETON ROBOT IN PATIENTS WITH TRANSIENT ISCHEMIC ATTACK

Background: Transient ischemic attack (TIA) is a condition causing focal neurological deficits lasting less than 24hrs. TIA patients present similarly to other conditions with rapid onset of neurological symptoms such as migraine. The accurate diagnosis of TIA is critical because it serves as a warning for subsequent stroke. Furthermore, cognitive deficit associated with TIA may predict the development of dementia. Therefore, characterizing the cognitive symptoms of TIA patients and discriminating these patients from those with similar symptoms is important for proper diagnosis and treatment. Currently the diagnosis of TIA is made on clinical and radiographic evidence. Robotic assessment, with instruments such as the KINARM, may improve the identification of cognitive impairment in TIA patients. Methods: In this prospective cohort study, two KINARM tests, trail making task (TMT) and spatial span task (SST), were used to detect cognitive deficits. Two study groups were made. The TIA group was tested at 5 time points over the span of a year. The migraine active control group had one initial visit and another a year later. Both of these groups were compared to a normative database of approximately 400 healthy volunteers. From this database age and sex matched normative data was used to calculate Z-scores for the TMT. The Montreal Cognitive Assessment (MoCA) was also administered to both groups. Results: 31 participants were recruited, 20 TIA group and 11 active controls (mean ± SD age= 66 ± 11.3 and 62 ± 14.5). There was no significant difference in TIA and active control group MoCA scores. The TMT was able to detect cognitive impairment in TIA and migraine group. Also, both KINARM tasks could detect significant differences in performance between TIA and migraine patients while the MoCA could not. Changes in TIA and migraine performance on the MoCA, TMT, and SST were observed. Conclusions: The robotic KINARM exoskeleton can be used to assess cognitive deficits in TIA patients.