Bing Speech Recognition

1. Install the skill in your ARC project and make sure it opens without errors.
2. Confirm your PC has internet access. Cloud speech recognition requires the audio to be uploaded.
3. Pick a microphone (USB headset, Bluetooth mic, webcam mic, etc.).
4. Test the microphone in Windows first (make sure the input meter moves when you talk).

1. Start Recording / Start Listening Button
Starts speech recognition. The skill begins waiting for speech, captures your voice, converts it to text, and shows the result in the Response Display. If you are brand new, click this first to confirm your mic works.

2. Audio Waveform
This is your quick microphone test. When you talk, you should see movement. If the waveform is flat, Windows is probably using the wrong microphone or the mic volume is too low.

3. Response Display (Log)
Shows what ARC recognized, plus silence detection and helpful status/diagnostic messages. Use this log to troubleshoot recognition quality, wake word confidence, and microphone settings.

Phrase List
A list of phrases you want the skill to recognize as “commands”. When your spoken words match a phrase, ARC can treat it as handled (a known command). Keep phrases short and distinct for best accuracy.
Examples: “robot start”, “turn left”, “stop”, “look at me”.

Not Handled Script
Runs when speech is recognized but it does not match anything in the Phrase List. This is useful when you want a fallback behavior (for example: “I didn’t understand” or send text to an AI). This script is skipped if a phrase match occurs.

All Recognized Script
Runs every time something is recognized, whether or not it matches your Phrase List. The recognized text is stored in a variable (default: $BingSpeech), so you can read it in your script.
Common use: send all speech to a chatbot, log everything, or do your own custom parsing.

Start Listening Script
Runs whenever the skill begins listening. This is commonly used for feedback such as: turning on an LED, showing a message on-screen, or playing a “listening” sound so you know it’s active.

Variable Field
The name of the ARC variable that will store the recognized text. Default is $BingSpeech. Because it’s a normal ARC variable, you can access it from:

• JavaScript or Python
• Other skills that read variables
• Scripts using GetVar()

Auto Record Using Wake Word
Enables wake-word detection (similar to “Alexa” / “Hey Google”). When the wake word is detected, the skill automatically starts listening for your command.
Beginner note: Wake words are convenient, but can be unreliable in noisy rooms. Push-to-talk is usually more dependable for robots.

Wake Word Sound
Plays a sound through the PC’s default speakers when the wake word is detected, so you know the robot “heard” the wake word.

Min Wake Word Confidence (0.0 – 1.0)
How “sure” the system must be before it accepts the wake word. Default is 0.75.

• Increase this if the wake word triggers by accident.
• Decrease this if you have to repeat the wake word too often.

Play Wake Word Sound with ControlCommand()
If enabled, the wake sound will also play when you start listening from a script using ControlCommand(). Useful if you want consistent audio feedback no matter how listening was started.

Stop Punctuation
Removes punctuation from recognized text. This can make command matching easier.
Example: “turn left.” becomes “turn left”.

Setup Microphone
Opens Windows audio settings so you can choose the correct input device and adjust levels. Use this if the waveform does not move when you talk.

Max Recording Length (Seconds)
Limits how long the skill will listen before it stops automatically. This helps prevent the robot from listening forever, which can reduce accidental triggers and excessive background audio uploads.

Language Drop-down
Selects the language/locale used for recognition (example: en-US). Available options depend on what is installed and supported on your system. If you don’t see the language you want, check Microsoft’s language/locale list and use the correct locale value: Microsoft Speech-to-Text language support.

1. Open the skill and click Setup Microphone. In Windows, select the mic you want to use.
2. Speak normally and confirm the input meter moves in Windows and the audio waveform moves in ARC.
3. Select the correct Language (for example: English (United States)).
4. Click Start Listening (or Start Recording) and say a short phrase like: “hello robot”.
5. Look at the Response Display and confirm your words appear as text.

What this script is doing

• Waits until the button is pressed
• Starts listening
• Waits until the button is released
• Stops listening, then the service converts the recorded speech into text

• Internet connection is required (speech recognition runs in the cloud).
• If your robot uses a separate WiFi network (for example, connecting directly to an EZ-B), you may need dual network connections (ex: Ethernet + WiFi, or two WiFi adapters). Learn more: Dual network connections.