Prosody Workstation—Version 2

Abstract

The Prosody Workstation, Version 2 (PW2) is a low-cost, PC-based system for interactively teaching and researching prosody. This device does four different, inter-related jobs. On output from the host PC, it (1) synthesizes model utterances from stored LPCs and prosodic control parameters. On user input, it (2) extracts the F0 and intensity of the user’s productions, for comparison with the model as presented. The host PC then (3) visually displays the original model and the user’s response, with similarities and differences highlighted for the user’s information. Finally, it (4) resynthesizes either the original model or the user’s response for auditory comparison. Whereas PW1 used an expensive and no longer available signal processing board, PW2 runs on any PC running Windows and with SoundBlaster compatible sound I/O. PW2 features of interest are (a) scoring algorithms and visual displays are user selectable, permitting (e.g.) time and/or F0 normalization, selective focus on specific prosodic elements, and use of different F0 scales; (b) data acquisition is efficient, with one F0 and intensity estimate calculated for each 10 or 20 ms frame; (c) data extraction and resynthesis routines are coded as C++ DLLs, while the command modules are visual basic, permitting efficient extension of functionality. [Work supported in part by NIDCD and Michigan State University.]