Thanks to the use of reliable speech recognition methods, most information from the system, i.e., the automatic scoring, the feedback instruction, and rhythm pattern templates, were proven to be highly reliable. Automatic training performance was achieved in nearly realtime. However, in acoustic analysis, a computational delay was pointed out as an improvement point, especially while calculating LPC spectrum parameters. At present, the CAPL system runs on Workstation, but it can be transferred to personal computers at the same level of performance.
