Speech Recognition

1. Pause Checkbox
Pauses phrase actions. While paused, the microphone keeps listening and the waveform keeps updating, but recognized phrases will not execute their scripts. The Enable and Disable phrases are the exception — they are always processed, which is what allows you to unpause the skill with your voice.
The pause state is saved with your project and restored when the project is loaded.

2. Phrase List Button
Toggles a window listing the phrases configured in the Settings, so you can quickly review what the skill is listening for. Clicking a phrase in that window triggers it exactly as if it had been spoken with 100% confidence, which is handy for testing your scripts without a microphone. Press Esc to close the window.

3. Audio Waveform
Displays live audio waveform feedback to confirm your microphone is configured correctly and actively receiving sound. The waveform is drawn in green when the skill is active and in red when it is paused.

4. Response Display
Shows detection and execution feedback. When a phrase is recognized, the log displays the matched phrase and its confidence in brackets, for example robot move forward (0.89). Other messages you may see:

• Low confidence: <phrase> (0.62) — heard, but below your confidence threshold.
• Enabled / Disabled — the Enable or Disable phrase was recognized.
• Weird, heard but not detected: <phrase> — the engine returned a phrase that is not in the current list (usually after the list was changed while listening).

5. Audio Quality Messages
When the recognition engine reports an audio signal problem, a short hint is displayed: No audio signal, Speak slower, Speak faster, Speak quieter, Speak louder, or Background audio too noisy. The message is cleared the next time a phrase is successfully matched.

The skill automatically pauses itself while ARC is speaking, so the robot does not recognize its own speaker output as a command. When speech begins, Pause is checked. When all speech has finished, the Pause checkbox is restored to the value you had set before the robot started talking — so if you had manually paused the skill, it stays paused afterward.

Overlapping or back-to-back speech is counted, so the skill only resumes after the last utterance completes. If a completion event never arrives (for example, speech that is rendered to a file or a stream that fails), a watchdog force-resumes after 60 seconds and notes the recovery in the Response Display.

1. Confidence Drop-down
Sets the minimum confidence required for a phrase to be accepted, selectable from 0.50 to 0.95 (default 0.75). Phrases detected below this threshold do not execute their action; they are logged as Low confidence and run the Low Confidence script instead. If your phrases are not being detected reliably, reduce the confidence setting.
Note: Lower confidence thresholds increase the chance of false positives.

2. Setup Microphone Button
Opens the Windows Recording devices dialog (Sound control panel). Use this to verify the correct input device is selected and to confirm the audio meter responds when you speak.

3. Recognition Scripts
All Recognized runs whenever a phrase from the list is matched at or above the confidence threshold. It runs before that phrase’s own script and completes first, so it is a good place for logging, diagnostics, or common setup shared by every phrase.
Low Confidence runs when a phrase was heard below the confidence threshold. It does not run while the skill is paused.

4. Enable / Disable Phrase Fields
Optional phrases that unpause and pause the skill by voice. They are added to the recognizer dictionary automatically and are matched even while the skill is paused, which is how the Enable phrase can wake it back up. Leave them blank if you do not want voice-controlled pausing. Matching is case-insensitive.

5. Enable / Disable Command Scripts
The scripts executed when the Enable/Disable phrases are recognized. This is commonly used to control robot behavior such as entering a “listening mode” or disabling speech during loud activities.

6. Language Drop-down
ARC uses the speech recognition capabilities built into Windows, so the list contains only the recognizer languages installed on your computer. ARC defaults to EN-US if installed, otherwise it uses the first installed language that supports speech recognition. If multiple supported languages are installed, select the desired language here.

For more information on installing additional speech recognition languages, view: https://www.tenforums.com/tutorials/120631-change-speech-recognition-language-windows-10-a.html

Here is how to add a new language pack:

1. Go to Start and open Settings.
2. Select Time & language > Language.
3. Select the language you want to add speech to, then select the Next button.
4. Select the speech options you want included with the language.
5. Sign out and sign back in for the new speech pack to be added to speech options.
6. Go back to Settings > Time & language > Language, select your new language, and move it to the top to make it default.
7. Go to Speech and ensure the Speech language setting matches your new default language.
8. Sign out and sign back in for the new settings to take effect.
9. Select the desired language from the ARC Speech Recognition configuration menu.

7. Confidence Variable
The global variable that stores the confidence value (decimal, 0 to 1) of the last recognized phrase. Defaults to $SpeechConfidence. The name must begin with a $.

8. Phrase Variable
The global variable that stores the text of the last recognized phrase. Defaults to $SpeechPhrase. The name must begin with a $.
Both variables are updated for every recognition — including low confidence results and while the skill is paused — so a script can inspect what was heard even when no action was executed.

9. Phrase List
The list of phrases to recognize. You can customize the defaults and add additional phrases. Rows with an empty phrase or an empty command are discarded when you press OK, and leading/trailing spaces are trimmed from phrases.

10. Command List
The command/action corresponding to each phrase in the same row. Each row can be an EZ-Script, Blockly, JavaScript, or Python script. You can customize commands and add additional rows.

11. List Management Buttons
Buttons for managing phrase rows: move up/down, insert, append, and delete.

When Windows reports a recognized phrase, the skill processes it in this order:

1. The phrase and confidence variables are updated.
2. If the confidence is below the threshold, the result is logged as Low confidence, the Low Confidence script runs (unless paused), and processing stops.
3. If the phrase matches the Enable phrase, Pause is turned off and the Enable script runs. This happens even while paused.
4. If the phrase matches the Disable phrase, Pause is turned on and the Disable script runs.
5. If the skill is paused, processing stops here — no phrase script runs.
6. The phrase list is searched (case-insensitive). On a match, the All Recognized script runs to completion, then the phrase’s own script is started.
7. If nothing matched, the log shows Weird, heard but not detected.

1. Install, configure, and test your audio input device (see Resources below).
2. Add the Speech Recognition skill to your ARC project: Project → Add Skill → Audio → Speech Recognition.
   Note: This is different from Advanced Speech Recognition.
3. In Settings, configure your phrases and the corresponding commands/scripts.
4. Save your settings, then speak your configured phrases into the microphone to trigger the actions.
5. Use the Phrase List button to test a phrase by clicking it, which runs the action without needing to speak.