Visual acuity test for isolated words using speech recognition

Visual acuity tests are performed by doctors to assess a patient's visual acuity. Health practitioners carry out this test manually on daily basis. This proposed technique aims at the ease of accurately testing vision anywhere instead of planning a visit to a practitioner. In this interactive method, a user utters isolated words as a guess input to the system from a table of selected words. The system accepts user's utterance in the form of speech and performs processing in two steps i.e. extraction of salient features or speech vectors from the speech signal sufficient enough to represent the utterance and application to compare the speech vectors extracted. The system will identify the correct and wrong guesses in parallel with each isolated word uttered by the user. Isolated words are determined by the system as part of the test. The feature sets extracted from the isolated words database includes mel frequency cepstral coefficients and computation of perceptual parameters which are classified by optimum class boundaries using Support Vector Machine. Challenging findings and evaluations are discussed.