Confusion Network Combination Approach For Improving Automatic Speech Recognition

This paper describes confusion network combination (CNC), which is capable of transforming a very large lattice into a confusion network. This method is one of the system combination techniques in word-level for automatic speech recognition and used HTK to create two systems (Mel-Frequency Cepstrum Coefficient and Perceptual Linear Predictive) for combination. Experiments performed on FARSDAT database show that the accuracy of automatic speech recognition systems can be improved by the combination of different confusion network. We suggest some directions that provide an absolute word accuracy rate (WAR) increment of about 1%