Abstract
In this paper, a Web application that allows people with speech impairments to train and amend pronunciation of words by repeating words presented in a interface with visual and auditory stimuli, is presented. With this app, we aim at present a training assistive tool to be used by the user, in a daily situation, where the main asset is the feedback about the rate of correct pronunciation, thus allowing users to train, autonomously, at home. Additionally, we present a preliminary user study to assess usability of the proposed application. For that, we invited a group of people with speech impairments to perform training tasks with the application. In the assessment, we register the following parameters: effectiveness (successful conclusion of the training tasks and percentage of pronouncing success); efficiency (number of errors made, number and type of difficulties found); and, satisfaction (the acceptance and comfort to complete the training tasks). Overall, the preliminary results showed that participants achieved an efficiency rate of 53.4% in the accomplishment of the proposed tasks. On average, for each image, participants missed at least 2 times, where the most registered difficulty was the pronunciation of words with accents. Regarding satisfaction, we noticed that they like it as they ask not only to repeat the test but also the addition of more pictures to perform more tests with the proposed web application.

Abstract
The continuous use of the muscles in any kind of physical exercises results in muscular fatigue, which can be defined as the incapability of the muscle to perform with the same effectiveness over the course of time. The analysis of physical exercise intensity has great importance in various fields, including sports and physiotherapy. In this paper, the rate of blinking eyes and the change in shape of mouth throughout the physical exercise are analyzed using computer vision techniques, and compared with the perceived exertion. The experiments were done using the facial video of three athletes, grabbed during a stationary cycle of physical exercise, until maximal muscle activity was achieved. The perceived exertion was reported at the end of each minute. The blinking of the eyes and opening of the mouth were detected by counting the number of bright pixels in the region of interest of an eye and of the mouth. These regions were detected using the Viola and Jones algorithm. We have proved the existence of a correlation between the opening and closing of the mouth and the eye-blinking rates with the physical exercise intensity (i.e., the higher the exercise intensity, the higher the rate of eye-blinking and mouth opening and closing). We obtained 95% accuracy in blinking eye detection.