High performance telephone bandwidth speaker independent continuous digit recognition

Abstract

The development of a high-performance telephone-bandwidth speaker independent connected digit recognizer for Italian is described. The CSLU Speech Toolkit was used to develop and implement the hybrid ANN/HMM system, which is trained on context-dependent categories to account for coarticulatory variation. Various front-end processing and system architectures were compared and, when the best features (MFCC with CMS + /spl Delta/) and network (4-layer fully connected feed-forward network) were considered, there was a 98.92% word recognition accuracy and a 92.62% sentence recognition accuracy on a test set of the FIELD continuous digits recognition task.