dtmf recognition accuracy with *, #, and digits
Posted: Tue Nov 16, 2004 10:36 pm
I would like to recognize commands of up to 3 keystrokes, with the possiblility of the first two commands as *.
I am using the following grammar:
<grammar type="application/x-jsgf" mode="dtmf">
["*"]["*"](1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 0 | "#")
</grammar>
The * commands are seldom recognized.
I swapped the role of the * key with 1 as follows:
<grammar type="application/x-jsgf" mode="dtmf">
[1][1]("*" | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 0 | "#")
</grammar>
This is very accurate and never misses. Do you have any suggestions to improve the performance of the first case?
Is it possible to allow bargein only with dtmf tones and to ignore all extraneous noise?
I am using the following grammar:
<grammar type="application/x-jsgf" mode="dtmf">
["*"]["*"](1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 0 | "#")
</grammar>
The * commands are seldom recognized.
I swapped the role of the * key with 1 as follows:
<grammar type="application/x-jsgf" mode="dtmf">
[1][1]("*" | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 0 | "#")
</grammar>
This is very accurate and never misses. Do you have any suggestions to improve the performance of the first case?
Is it possible to allow bargein only with dtmf tones and to ignore all extraneous noise?