Open Speech Command Translator System for Executing Positive, Negative, and Neutral Decisions

Abstract

Abstract - This research proposes a system that uses an open speech command and translates it into a Positive, Negative, or Neutral decision. An ordinary voice command targeted to an object is generally composed of two parts: the object identification, and the desired instruction to execute from it. Using this structure, several voice features could be analyzed such as pitch, intensity, formants, among other statistical characteristics to recognize if a pattern exists. For this research, the “object” segment was ideally converted to text using existing speech-to-text tools; on the other hand, the “instruction” segment was ideally processed by applying a Naïve Bayes Classifier to classify instructions as Positive, Negative, or Neutral with up to 72% accuracy. This idea could be used in external systems for people with special needs, among other yet unimagined applications.The system will enable improved natural interaction between humans and computers by using a basic but effective open communication. Key Terms - Command, Decision, Language,Translator