Um estudo sobre a eficácia de uma abordagem híbrida para reconhecimento de voz

Abstract

In the last decade, there has been significant progress in voice recognition technologies with applications in Intelligent Environments (AmI). The possibility of using the voice as a form of interaction enables people to integrate the scenario that some natural and transparent manner. Entreated, there are still obstacles to be overcome, such as the treatment of general expressions, variations according to the profile of users, voice timbre and spontaneous behavior. To contribute to solving some of these problems, this work proposes a model to allow natural interactions via voice, including a hybrid and adaptive approach to profile users. To validate the model, a platform capable of interpreting frequently used commands was created in a local structure, and new commands in a cloud structure. According to usability testing, it was observed that there are significant gains to the hybrid approach, allowing adaptation of the system to the user’s profile.