资 源 简 介
The speech recognition diagnosis is the code participated to my TSD 2010 submission. The key idea is linking dynamic program
events (program spectra) with erroneous program behaviour at the level of code lines, and localize the lines that match error indicators most closely. Diagnostic accuracy is then computed to rate the finding for a human examiner. The diagnosis is shown at a toy experiment, based on Torch3 free ASR and free OLLO speech database.
