Speech Recognition Mic Variable

Question

, I'm using the Speech Recognition Robot Skill in ARC. From a Python script running in that Speech Recognition Skill, I'm attempting to start Object Recognition in the Camera Robot Skill. I converted the EZ-Script code from the "Object recognition" example by DJ Sures to Python. The script executes, but it does not capture the speech text from the user microphone in response to the prompt "What is the name of the Object?" It seems the Speech Recognition Skill does not expose a variable equivalent to the EZ-Script example's "$BingSpeechPhrase". I have tried using "SpeechPhrase", but that always contains the previously spoken robot voice (TTS), NOT the text captured from the user microphone. Is there a variable I can use to obtain the captured speech text from the Speech Recognition Skill? Note: I have a bespoke variable named "$CurrentSpeechPhrase" that another script uses to manage the robot voice and jaw animations. My Python script (converted from EZ-Script) is shown below: python import time # Ask for the object name setVar("$CurrentSpeechPhrase", "what is the name of the new object") setVar("$IsSaying", 1) while getVar("$IsSaying") == 1: time.sleep(3.0) # Reset speech variable and start listening setVar("$SpeechPhrase", "") controlCommand("Speech Recognition", "PauseOff") time.sleep(0.25) setVar("$SpeechPhrase","?????????") # # Confirm object name with the user first test object_name = getVar("$SpeechPhrase") # This does not match the input from the microphone # setVar("$CurrentSpeechPhrase", "I heard you say " + str(object_name)) setVar("$IsSaying", 1) while getVar("$IsSaying") == 1: time.sleep(0.1) # Wait for Bing Speech to get a response while getVar("$SpeechPhrase") == "": time.sleep(0.25) # Confirm object name with the user object_name = getVar("$SpeechPhrase") setVar("$CurrentSpeechPhrase", "Excellent. Let me know when you are ready to teach me what the " + str(object_name) + " is.") setVar("$IsSaying", 1) while getVar("$IsSaying") == 1: time.sleep(0.1) # Wait for 'OK' or 'Cancel' voice confirmation (Timeout: 60 seconds) # Note: waitForSpeech is a native ARC global function resp = Audio.waitForSpeech(60, "OK", "Cancel") if resp != "OK": setVar("$CurrentSpeechPhrase", "ok, I will not learn the " + str(object_name)) setVar("$IsSaying", 1) # Stop script execution exit() # Disable current tracking to prepare for training controlCommand("Camera", "CameraObjectTrackingDisable") # Train the new object controlCommand("Train Vision Object By Script", "Learn", str(object_name)) time.sleep(1.0) # Wait while the camera is actively learning while getVar("$CameraIsLearning") == True: time.sleep(0.25) # Success confirmation setVar("$CurrentSpeechPhrase", "Great, I know what a " + str(object_name) + " is.") setVar("$IsSaying", 1) while getVar("$IsSaying") == 1: time.sleep(0.1) # Re-enable regular camera tracking controlCommand("Camera", "CameraObjectTracking")  Thanks.

Athena · Accepted Answer

Yes - the Speech Recognition skill does not provide a "$BingSpeechPhrase"-style variable for the recognized user phrase in the way the old Bing speech examples did. What you want is usually the last recognized phrase variable exposed by the Speech Recognition skill itself, which in ARC is typically: - $SpeechPhrase = the last recognized phrase - $SpeechConfidence = confidence of the last recognition However, in your case the reason you’re seeing the robot’s TTS text instead of the user’s microphone input is likely because your own custom variable ($CurrentSpeechPhrase) and/or the speech/jaw logic is updating $SpeechPhrase or your script is reading it at the wrong time. ## Important distinction - Speech Recognition skill listens to the microphone - Audio.say... / TTS is the robot speaking - Those are separate paths So if "$SpeechPhrase" is showing the spoken robot text, then that variable is likely being used by another script as a shared "current phrase" variable, not the Speech Recognition skill’s recognized input. --- ## Better approach Use the Speech Recognition skill’s built-in recognized phrase variable from that control, not your custom phrase variable. If you’re using the Speech Recognition robot skill, check its settings: - Open the skill’s kebab menu (three lines in the top-right of the skill) - Look at the Phrase Variable setting - Also check the Confidence Variable You can point those to unique variable names, for example: - "$RecognizedPhrase" - "$RecognizedConfidence" That avoids collisions with your TTS/jaw script. --- ## Recommended fix for your project ### 1) Change the Speech Recognition skill variables In the Speech Recognition skill settings: - Set Phrase Variable to something unique like: - "$RecognizedPhrase" - Set Confidence Variable to something like: - "$RecognizedConfidence" ### 2) Read that variable in Python Then your script should use: python object_name = getVar("$RecognizedPhrase", "") print(object_name)  --- ## Why your current script is failing This line is the key issue: python setVar("$SpeechPhrase", "")  If "$SpeechPhrase" is already being used by another part of your project, clearing it right before listening can cause conflicts. Also, your jaw/TTS logic is likely writing to a phrase variable that is not the microphone result. --- ## Suggested Python pattern Here’s a cleaner version of the speech capture portion: python # Clear only the dedicated recognition variable setVar("$RecognizedPhrase", "") # Enable speech recognition if needed controlCommand("Speech Recognition", "PauseOff") # Wait until a phrase is recognized while getVar("$RecognizedPhrase", "") == "": sleep(100) object_name = getVar("$RecognizedPhrase", "") print(object_name)  If you need to wait for a specific prompt response like OK or Cancel, you can do: python resp = Audio.waitForSpeech(60, "ok", "cancel") if resp != "ok": setVar("$CurrentSpeechPhrase", "ok, I will not learn the " + str(object_name)) exit()  --- ## What I’d recommend in ARC Since you have multiple scripts interacting, the safest setup is: - Use $CurrentSpeechPhrase only for robot speech / jaw animation - Use $RecognizedPhrase only for microphone recognition - Use $RecognizedConfidence only for confidence filtering That separation will prevent the two systems from stepping on each other. --- ## Useful links - Speech Recognition skill: https://synthiam.com/Support/Skills/Audio/Speech-Recognition?id=16118 - Configuring robot skills: https://synthiam.com/Support/ARC-Overview/robot-skills/configuring-robot-skills - ControlCommand documentation: https://synthiam.com/Support/Programming/control-command --- If you want, I can help you rewrite your whole script into clean ARC Python so it: 1. prompts for the object name, 2. captures the mic phrase correctly, 3. confirms it, 4. then starts the camera learning skill.

Cosgrove · Answer

@Athena, thanks, yes a rewrite will save me time. Thanks. I did not know that there was a variable = $RecognizedPhrase One question: Can I still use $SpeechPhrase in other scripts or do I now have to change all other uses to $RecognisedPhrase as that would seem to just transfer the clash of variables to $RecognisedPhrase. Thanks again

Cosgrove · Answer

@Athena, @Synthiam Support, the problem remains. The Speech Recognition Skill does not populate the variable $RecognizedPhrase with the input from the microphone. By the way, some of the code you provided caused errors in Python - this line while getVar("$RecognizedPhrase", "") == "": results in an error "invoke(requires one argument not 2". This Python code runs but does not give the intended result. python import time # Reset Speech variable and start listening # Clear only the dedicated recognition variable setVar("$RecognizedPhrase", " ") # Enable speech recognition if needed controlCommand("Speech Recognition", "PauseOff") # Wait until a phrase is recognized while getVar("$RecognizedPhrase") == "": #time.sleep(0.25) object_name = getVar("$RecognizedPhrase") print(object_name) # Confirm object name with the user first test object_name = getVar("$RecognizedPhrase") setVar("$CurrentSpeechPhrase", "I heard you say " + str(object_name)) setVar("$IsSaying", 1) while getVar("$IsSaying") == 1: time.sleep(0.1) # Wait for Speech to get a response while getVar("$RecognizedPhrase") == "": time.sleep(0.25) # Confirm object name with the user object_name = getVar("$RecognizedPhrase") setVar("$CurrentSpeechPhrase", "Excellent. Let me know when you are ready to teach me what the " + str(object_name) + " is.") setVar("$IsSaying", 1) while getVar("$IsSaying") == 1: time.sleep(0.1) # Wait for 'OK' or 'Cancel' voice confirmation (Timeout: 60 seconds) # Note: waitForSpeech is a native ARC global function resp = Audio.waitForSpeech(60, "ok", "cancel") if resp != "ok": setVar("$CurrentSpeechPhrase", "ok, I will not learn the " + str(object_name)) exit() # Disable current tracking to prepare for training controlCommand("Camera", "CameraObjectTrackingDisable")

Synthiam Support · Answer

There is no such variable as the one you commented on. The Speech recognition robot skill refers to the recognized phrase as $SpeechPhrase However, speech recognition only recognizes the phrase configured in the list. Because of that, you should only have the scripts associated with the respective phrase. Athena can explain more. If you want free text for any phrases, use a different speech recognizer.

Cosgrove · Answer

@Synthiam Support. I'm not happy with your response, given that @Athena suggested code that might work and unfortunately it did not. I think the AI expected it would (as I did). Here's my take on what you have said - Phrases entered in the "Phrase list" are attached to the SpeechPhrase variable (or whatever variable you put in the box on the first screen in the Skill), but once a script starts execution, there is no recognition of Microphone input (this is as you said line 2 above). Bing Speech recognition does this - attaches input to a variable but the in-house Speech recognition does not. Why not? One might think that as Bing Speech is restricted in use unless you are a PRO user, Synthiam have limited functionality in the Speech Recognition Skill to have makers use Bing with the 10/day phrase restriction or subscribe to PRO and use Bing. I'd like this question marked unsolved.

Synthiam Support · Answer

The speech recognition uses the Windows built-in speech recognition system, which requires pre-defined phrases. It does not convert "any spoken words into text". The detected phrase is stored in the variable mentioned above. Bing speech recognition uses a third-party paid service that converts any spoken words into text. In 99.9% of cases, the variable of the Speech Recognition robot skill is not quite useful because the recognized phrase executes the respective code associated with it. For example... Phrase: "Say Hello" will execute the script associated with that phrase. Phrase: "Move Forward" will execute the script associated with that phrase. Hope that helps! Read more here: [opt:skill:16118]

Synthiam Support · Answer

Oh, and one more thing. We attempted to make a free version of a full speech recognition system using the "built-in speech recognition," and it is called Total Speech Recognition. Because localized speech recognition technology is very limited, it is incredibly unreliable, unfortunately. You can find more about it here: [opt:skill:17688]

Cosgrove · Answer

@Synthiam Support - Thanks. I guess I should direct my complaint to Bill Gates!

Synthiam Support · Answer

Hahaha - that would be a good start! The biggest challenge is that all these great services have turned into paid online subscriptions. None of these services publish their product for free because they need to generate revenue to pay for their employees. We are always keeping our eyes open for open-source products as well. Still, for services involving recognizers and synthesizers, they seem to be limited to large datacenter-style computing, probably because they require so much processing to convert any speech into words.

If you use Total Speech Recognition, make sure you use a very good microphone. Generally, we recommend a headset microphone. Or, one of those microphones that you can wear on your lapel.

Cosgrove

Speech Recognition Mic Variable

Upgrade to ARC Pro

Products

Community

Support

About