So I managed to accomplish something in like 2 days that I couldn't do after a year in Japan doing a undergrad research project. I threw together a voice interactive application. Granted the machines were much slower back then, and I had only an old version of the HTK toolkit and some very bad training data to work with back then. Alcohol and Karaoke may have also been a factor. And considering I was a good bit lower down in the recognition process (tagging my own speech recordings for training, generating the HMM bigram / trigram models for comparison testing, etc...) today was a bit anti-climatic. Also, I haven't really made much here.. just mashed two awesome projects together.
Continue reading Voice interactive system.