SPEECH RECOGNITION DEVICE

PURPOSE:To enable a recognition as exact speech data by providing a speech pattern dividing process section using a neural network and a standard pattern dividing process section. CONSTITUTION:This device is provided with a feature extracting section 101 which extracts the features of the speech data inputted and constructs a speech pattern corresponding to the speech data, a storage section 102 which stores a standard pattern that produced based on a learning pattern and a feautre matching section 103 which calculates the degree of similarity between the voice pattern and the standard pattern stored in the section 102 and outputs the recognition results when a high-degree similarity is calculated. Moreover, when a low-degree similarity is calculated by the section 103, a speech pattern dividing process section 104 which repeatedly executes dividing processes of the speech pattern using a neural circuit network in a prescribed from and a standard pattern dividing process section 105 which performs dividing processes of the standard pattern using the neural network in a prescribed form are provided.