Setup Sphinx
When you install Simon, it has a daemon Simond, this daemon controls the model compilation. This daemon has a fronted:ksimond
Open it up and go to "Speech Model Compilation"
![](simond_empty.png)
As you see there are some paths missing. Get the paths from your Distribution package manager from the SPHINX package and fill the gaps!
Here on Gentoo it looks like this:
![](simond_filled.png)
Now on to the Simon config.
Basic Simon configuration
Here are some remarks about my Simon config, I found that they work best for me.
There is no guarantee that this will work with your setup. Every hardware is different! Read the Handbook from Simon!
Go to "Settings" -> "Configure Simon" -> "General"
Tick box: "Show warning when Simon detects a problem with a sample"
Then go to "Recording" -> "Voice Activity Detection"
Use this settings.
- Cutofflevel: Everything below this level is considered “silence” (background noise).
- HeadMargin: Cache for as long as head margin to start consider it a real sample. (gives better training results as longer samples yield better recognition results)
- Tail Margin:After the recording went below the cutoff level. Simon waits that long. See reason above.
- Skip Samples: Skip samples shorter than...
HeadMargin: 1350
Tail Margin: 1300
Skip Samples: 150
Then go to "Training"
Signal Noise Ratio: 300-500 for me.
Depends on the environment you are in. Lower means more background noise!
Tick box: Volume calibration
Tick box: Power Training (not recommended at the beginning!)
<< go_back -Main Site- >> Linjark modules