Welcome to the world of Scilab Speech

fAQ

1. Why SciSpeech can not work correctly?

Please note that our software must installed under the directory of scilab-4.0, for example: "C:\Program Files\scilab-4.0"

2. Why we can only load the "SciSpeech\doc\sample.wav" when select
Analysis-->Pre-Emphasis?

One of the functions is to filter the noise in the audio file, and the rate of the audio file that we provide is 8000HZ, so you have to load an audio file whose rate is also 8000HZ, here we only provide sample.wav.

3. What do Synthesis-->Psola method do?

Here we implement the PSOLA algorithm, and adjust the prosodics. In the GUI window we only use number 0-9 as samples. For more details, please refer to our document.

4. In the recognition process, why the recognition rate will be low and there will be error sometimes?

First of all, our recognition system is talker-dependent recognition system, users have to create their own model so that our system can recognize you speech command correctly.

Secondly, the reason that why there will be error is that the noise may be too much when you record, so we recommend you use a good microphone to record speech command.

Finally, we recommend you use the 16000HZ in the RecordSetting, if the sample rate is smaller than 16000HZ, then it is possible that the recognition rate will be low.

5. If I click pause botton and then click play, why it play the audio file from the start point?

If you want to play from the stop point, you have to click pause botton again.

6. Why the system do nothing when I click ZoomIn?

You have to select a section of wave first and then click ZoomIn botton.

7. Why it is quite slow when select Analysis-->WRLS-VFF?

This procedure include a lot of scitific computing, and there is also a 3D graphic, so this process will take a while, please be patient.

8. In the recognition process, why some of the command will be missed?

Please note that our system is single-word recognition system, so you have to stop about half a second when you record speech command. For more details, please refer to our demo.