This paper describes and examines efficient features for distributed automatic speech recognition (ASR) systems. The features are extracted by means of transformations of codecs parameters. These codecs are the ones used in cellular mobile networks and IP networks, which are the main applications in distributed speech recognition. In particular, features obtained from LPC and LSF parameters, in intervals of 10 ms and 20 ms, are analyzed in an HMM-based speaker independent recognizer. Moreover, we examine the best domain to perform features interpolation in distributed automatic speech recognition systems.
Distributed Speech Recognition; LSF; HMM
