Reconocimiento de voz

Abstract

This is a project about speech recognition, had been developing with de objective to establish a precedent to be utilizable on future investigations, like a document of reference and support. Base on two main points: a theory investigations and an implementation of an algorithm to recognition of isolated words. The investigation part is about speech technologies, speech signal processing and a DSP study, this ending because is an important tool in a real implementation. On the implementations, the work was base on two techniques. The first implementation characteristic was the use of DTW algorithm, and the second one used a neural network. The system with the first technique is based on two phases, the training phase, where the database with the words to recognition is created, and the recognition phase itself. The Neural Network system was more flexibility and with less functional blocks, which makes implementation simplify. Without requiring a database this system is easy to training for the recognition of different groups of words.