ADAPTATION OF AN ISOLATED WORD SPEECH RECOGNITION SYSTEM TO CONTINUOUS SPEECH USING MULTISECTION LVQ CODEBOOK MODIFICATION AND PROSODIC PARAMETER TRANSFORMATION

An improved, phoneme-based IWSR system is described, which employs a r obust reference data extraction procedure and achieves increased recog nition accuracy. Furthermore, a novel method for the adaptation of the IWSR-system to continuous speech is presented. The IWSR system employ s a multisection codebook design technique and the LVQ algorithm, whic h provide well-defined and accurate codebooks, minimize the influence of the within-word coarticulation and allow the use of time-sequence i nformation at the recognition stage. The adaptation method is based on modifications of the system's reference data codebook using a small a mount of representative continuous speech data and on linear transform ations of the main prosodic parameters (i.e. energy and duration). Ext ensive testing under different conditions (speaker dependent versus sp eaker independent reference data, single versus multisection codebooks , adapted versus unadapted codebooks, phoneme versus word recognition accuracy, etc.) has shown the efficiency of the proposed methods.