Hi there, I want to create an application with some word recognition in Slovak language. All I need is to recognize few isolated words (10-20). I've tried to map English phones to our words but the level of recognition is poor. I've read a lot of posts in this forum but still don't know how to improve the recognition.
Can anyone please help me?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Here's my problem. My application move the mouse via voice. It's designed for people with disabillities... Let's say they can't use their hands. You advise me to record the voice and recognise the file afterwards. The best way I can manage to do that is to signalise the beginning and end of recording (press ENTER...) to record spoken command. But how can I use this in my application? Users can use only speech to interract with computer.
This is little bit off topic, but please. Any advice is needed :-)
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hm, no actually there is easier way as I see now. Just replace in the LiveCMN with BatchCMN in the config file of HelloWorld and everything should be ok.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Trunk has few issues recognizing from the microphone on Windows. Try to record to a file and use WavFile as a base. Share the files you've made. I think it's quite easy to implement what you want.
Russian voxforge model could be closer to Slovak btw.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
"Try to record to a file and use WavFile as a base"
How am I supposed to understand this? Instead of recognising the input directly from microphone I shoud firstly record from microphone to WavFile and then recognise that file?
And another question.. I can´t find the way how to switch between recognising connected and isolated words. Can you help me with that?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi there, I want to create an application with some word recognition in Slovak language. All I need is to recognize few isolated words (10-20). I've tried to map English phones to our words but the level of recognition is poor. I've read a lot of posts in this forum but still don't know how to improve the recognition.
Can anyone please help me?
>It depends on your grammar you are using. If your jsgf has loops, it will be continuous recognition otherwise it's a single word recognition.
I should read the documentation for JSGF before asking... but I thought that this decision is made somewhere else...
Thank you Nickolay :-)
Going to try recording before recognising...
Here's my problem. My application move the mouse via voice. It's designed for people with disabillities... Let's say they can't use their hands. You advise me to record the voice and recognise the file afterwards. The best way I can manage to do that is to signalise the beginning and end of recording (press ENTER...) to record spoken command. But how can I use this in my application? Users can use only speech to interract with computer.
This is little bit off topic, but please. Any advice is needed :-)
Or let's upgrade to latest svn trunk and fix all issues we have. Everyone will benefit from that :)
Hm, no actually there is easier way as I see now. Just replace in the LiveCMN with BatchCMN in the config file of HelloWorld and everything should be ok.
I forgot to mension that I'm using SVN version of Sphinx 4 on Windows...
Trunk has few issues recognizing from the microphone on Windows. Try to record to a file and use WavFile as a base. Share the files you've made. I think it's quite easy to implement what you want.
Russian voxforge model could be closer to Slovak btw.
"Try to record to a file and use WavFile as a base"
How am I supposed to understand this? Instead of recognising the input directly from microphone I shoud firstly record from microphone to WavFile and then recognise that file?
And another question.. I can´t find the way how to switch between recognising connected and isolated words. Can you help me with that?
> Instead of recognising the input directly from microphone I shoud firstly record from microphone to WavFile and then recognise that file?
Exactly
> can´t find the way how to switch between recognising connected and isolated words. Can you help me with that?
It depends on your grammar you are using. If your jsgf has loops, it will be continuous recognition otherwise it's a single word recognition.