Hello, dear Android developers.
I'm an Android developer myself.
I'm developing some app related with speech recognition. This app has to recognize few voice commands (not much, 10-15 commands).
I chose the pattern similar to "Ok Google" style, i.e. user has to give some initial command and then app starts to recognize other commands.
I faced the problem which I'm not able to solve. Recognition of the rest of commands is not perfect but acceptable. But I can't reach the acceptable level of the initial command recognition. I got two extremes - in one case the app gets to many false actuations, in other case it's to hard to get true actuation. I can't reach reasonable ballance.
I'd like to meet the developer who have the experience with the speech recognition, who've already faced with the similar task and was successful with it. I'd like to get some sample app which demonstrate the acceptable level of the initial command recognition. When I say "acceptable" I mean - rare false actuations and at least 4/5 correct true actuations.
I don't have any requirement to the UI design of the sample app.
I'm using Pocketsphinx speech recognition engine in my app. But feel free to use any engine you think could be more effective.
May be you could suggest me more effective pattern without the initial command - I'd love to discuss it.
Thank you for your bids.