Today, we tested the SDK of the Hkust Flash.
Several problems were encountered.
Question 1
I used the speech recognition demo, and when I changed a voice file, I recognized his results as such.
confidence= grammar=0ID=nomatch name=nomatch:out-of-voca input=nomatch:o Ut-of-voca
After finding it, you need to modify the contents of this open grammar file.
fp = fopen ("gm_continuous_digit.abnf""RB"); if (NULL = = fp) { printf ("\nopen Grammar file failed!\n"); Goto grammar_exit; }
This fast-fly SDK can only be identified by a grammar file. Otherwise, it's not recognizable.
Question 2
Next I succeeded in identifying other example files. I started writing my own grammar file and started recording myself.
My grammar file
1.0 Gb2312;mode voice;language zh-= $yidong $digit <1-9>= Forward | 123456789 0;
Now I'm starting recording, and my recording command is
5 16000 -D PLUGHW:0 Test.wav
However, there is still a problem starting to recognize it.
10119
Officials say the problem is a format mismatch. But my format is perfectly correct.
Finally, I found that my voice was too light, and he recognized that it was noise, before the problem arose. Then I increased the volume.
end of speech recognition ============================================================= confidence= grammar=04 m confidence= grammar=0541 m confidence= grammar=0542 m ====== ======================================================= Press any key to exit ...
Flash SDK for Linux first experience