0

I'm trying to build a voice assistant. I'm using speech_recognition library for the same. Below is my code:

def takeCommand():
print("taking command")
r = Recognizer()
m = Microphone()

with m as source:
    
    try:
        print("speak...\n")
        r.adjust_for_ambient_noise(source=source, duration=0.5)
        audio = r.listen(source=source, timeout=5)
        command = None
        command = r.recognize_google(audio_data=audio).lower()
        print(command)
        # exporting the result 
        with open('recorded_test.txt',mode ='w') as file: 
            file.write("Recognized text:") 
            file.write("\n") 
            file.write(command) 
            print("ready!")
        return command.lower()

    except Exception as e:
        print("command could not be parsed")
        print('Exception: ' + str(e) )
        return None

Below is the code for the conversation function which I wrote for my assistant to give appropriate answers to my query:

def conversation():

while True:
    question = takeCommand()
    if question is not None:
        print("entered if, command not empty")
        if "thank you" in question and "jarvis" in question:
            speak("shutting down, sir!")
            break  
        elif "goodbye" in question and "jarvis" in question:
            speak("shutting down, sir!")
            break
        elif question:
            # some code to run appropriate function to accomplish the task
    else:
        speak("I'm afraid I couldn't recognize what you said. Please say again!")
            continue

I'm calling takeCommand() function inside my conversation() function. The thing which I want to accomplish here is that if the user is giving some command then the flow of control of my conversation function should be the same as per the if-else statements.

But, if the user is voluntarily choosing not to give any further commands then instead of running else in the conversation statement, the flow should go back to the main() from where I called the conversation() function.

So basically the code should understand on its own when the user is giving commands and when it's not.

Proposed Solution: I'm thinking of accomplishing this by running two functions in parallel to each other. One function will continuously take the audio input. In the meanwhile, other function will process it with packages like librosa or sounddevice to check the decibel levels of the audio input. if the levels are above certain threshold then the recorded audio should be sent to get recognized with speech_recognition library otherwise not.

Can anyone help me with accomplishing the above task?

It doesn't need to align with my proposed solution. If there is better way to accomplish this then do tell. I'm thinking ML and AI but don't know how.

desertnaut
  • 57,590
  • 26
  • 140
  • 166
styles
  • 7
  • 5

1 Answers1

0

one solution can be calculating rms using audioop against a threshold. If the value is above that threshold you can proceed with recording else return to checking the rms value.

link

Shalini Krishna
  • 51
  • 3
  • Can you share some sources or codes on how to do that? – styles Mar 23 '23 at 17:34
  • it'd be very helpful if u include code example please @Shalini Krishna – styles Mar 23 '23 at 18:27
  • check this link for rms calculation using audioop [link](https://docs.python.org/3/library/audioop.html#audioop.rms). Other solution would be using voice detection detection - https://github.com/wiseman/py-webrtcvad @styles – Shalini Krishna Mar 24 '23 at 06:11
  • Your answer could be improved with additional supporting information. Please [edit] to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers [in the help center](/help/how-to-answer). – Community Mar 28 '23 at 10:08
  • I have included the link in the above answer. I have tested this on ubuntu system. worked for me. let me know. @styles – Shalini Krishna Mar 28 '23 at 13:31
  • Thanks a lot, @ShaliniKrishna!! The resources shared by you were helpful. But I have a few other issues too while implementing your solution. Can we discuss this over discord voice chat or something? – styles Apr 03 '23 at 10:26
  • Sure, you can email at shalinikrishna67@gmail.com. Also if you find the solution usefule, do vote. – Shalini Krishna Apr 06 '23 at 10:17