i installed pocketsphinx on debian machine official repositories. tried train default acoustic model voice (indian english). results highly inaccurate.
1) can please suggest should done improve accuracy?
2) there better alternative in open source world speech recognition?
according google next conference on last wednesday opened speech text apis public using supporting of 80 language , work application in real-time streaming or batch mode .check article, recommend use return accurate results pocketsphinx .or check microsoft oxford project solution
Comments
Post a Comment