DEMONSTRATION - SPEECH RECOGNITION

Let's demonstrate speech recognition system for Japanese.
Here we distribute the sample application of Julius, the speech recognition engine that our laboratory develop and release, for Windows. It recognizes Japanese station names. Recognizing input Japanese speech, it displays the optimal station name in 8733 candidates.

System Operating Environment

• OS: Windows XP or Windows Vista
• Machine: Sound card that can record at 16-bit
• Other: Microphone

How to Use

Download

Download from the following link.

recognition_demo-0.5.zip

Extraction

The file is zipped. Right-Click on it and select "Extract All" to unzip it.

Execution

Open up the extracted folder and double-click on "recognition_demo(.bat)".

After the window is displayed, turn on the microphone and speak the Japanese station names. Was it recognized well?

Tips

• Speak clearly and slowly. But no need to speak separately individual hiragana.
• Adjust the volume of microphone properly. Refer to "Setting of Microphone Volume" below for the setting method.
• Execute it with little background noise.

Setting of Microphone Volume

Right-click on the volume icon on the bottom-right of the desktop, and select "Open Volume Control".

Select "Options" > "Properties" on Volume Control.

Select "Recording" in "Adjust volume for". Then in "Show the following volume controls", check "Microphone" and click on "OK".

Check "Select" in "microphone", adjust the volume.