I was also looking at the getBestTimedResult function. But it doesn't display any output as such. Are there any specific variables to be set in the config file to get the timing information.
Thanks,
Praveen
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I should elaborate. If the timestamps aren't consecutive, you know there is a gap, and the duration is of course trivial to calculate. What you can't tell from this basic information is how many non-recognitions happened within this gap. So the question is "does sphinx know it found something that it couldn't recognize, and does it maintain any information about each anomaly?"
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
> So the question is "does sphinx know it found something that it couldn't recognize, and does it maintain any information about each anomaly?"
It's better to read an introduction into HMM speech recogntion, like the tutorial in HTKBook or any other text book on ASR to avoid questions like this one. It will be helpful for you to understand how things work internally.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi All,
I tried out few examples of sphinx and found it really cool.
I was wondering whether sphinx also provides timeline info for audio detected?
ex: it detected word "green" at 3 secs on the timeline
Thanks
-Khan.
Can this function be called on unrecognized words or non-speech elements? For example, if the utterance in question is:
"She sells sea shells by the sea shore."
and Sphinx only transcribes:
"sells shells by the shore"
can Sphinx tell you that it skipped something and relay the timestamps of the gaps?
Yes, it's possible to get times. See Result.getBestTimedResult and search forum for examples.
I was also looking at the getBestTimedResult function. But it doesn't display any output as such. Are there any specific variables to be set in the config file to get the timing information.
Thanks,
Praveen
I should elaborate. If the timestamps aren't consecutive, you know there is a gap, and the duration is of course trivial to calculate. What you can't tell from this basic information is how many non-recognitions happened within this gap. So the question is "does sphinx know it found something that it couldn't recognize, and does it maintain any information about each anomaly?"
> So the question is "does sphinx know it found something that it couldn't recognize, and does it maintain any information about each anomaly?"
It's better to read an introduction into HMM speech recogntion, like the tutorial in HTKBook or any other text book on ASR to avoid questions like this one. It will be helpful for you to understand how things work internally.