CONNECTIVE RECOGNIZER AND METHOD

摘要

PURPOSE: A connective recognizer and method are provided to recognize an input sound which continues more than two words by performing one-pass algorithm. CONSTITUTION: A sound input device(10) receives and outputs a sound of a speaking person. A ROM(20) stores a sound detecting reference models. When a continuous sound from sound input device(10) is inputted to a CPU(30), the CPU(30) selects a reference model which corresponds to the input sound among reference models stored in the ROM(20) by performing one-pass algorithm and outputs the selected reference model as a sound detecting result. An object controller(40) controls an operation of a predetermined object according to a command for the continuous input sound detected by the CPU(30).