Can voice recognition module V3.1 recognize 3 different voices?
link to the module you are referring to?
have you read the spec?
if it's the ELECHOUSE Voice Recognition Module then it's speaker dependant.
We are 3 in our group, and I want my group mates also speak to voice recognition, but we are using 1 command only, it is possible?
you could may be cheat. Each command needs to be trained before the system can work. Say you want to be able to tell the module "countdown" ➜ instead of creating only one command for this, you create 3 different ones for the same word and each of you train one instance.
To the module they look different because the training data will be different, but the command will be recognised and then it's up to you to say in the code that regardless if the module recognised command1, command2 or command3 they are all the same and mean "countdown"
Spam and associated replies deleted.
And how can I fix this problem:
When I say a command it takes 3 -4 times before it recognized, what can I do?
Those modules do not work very well, and there is nothing you can do to fix them.
The module is not really good so don’t expect miracles
Choose words carefully so that they are easily differentiated and train the system in the environment it will be used
Okay, thank you.
Thanks
What should I do to recognize my voice easily. My problem is when I say "gobbledygook" as a voice coxmmand, it takes how many try to be recognize. What should be the solution for it? I am using Voice Recognition Module V3.1 by Elechouse.
Voice recognition is one if the hardest things in electronics to do. Even the big corporations have problems with it and have to send the information down to large computers to do this.
There are no easy answers.
Sorry.
Some smartphones like the iPhone do edge speech recognition with a speaker independent model built-in. To do this, the iPhone includes a dedicated super fast multicore neural network hardware to process the sound which is sampled from a finely tuned array of physical microphones and beamformers.
You just don't get that in Elechouse's hardware.
things I've seen matters a bit:
- train the model in the environment you are using it.
- Some "sounds" might be more difficult to distinguish than others due to microphone quality and environment. Try with a longer command
side note: what's gobbledygook ? moderator edit: redacted to be less spammy
gobbledygook is one of a kind medicine.
moderator edit: redacted to be less spammy
Building a drug dispenser?
Yes
Can you help me?
You need a very high quality, essentially 100% accurate voice recognition system for dispensing drugs. The Elechouse module is far too unreliable.
I cannot. Speech recognition is far outside my wheelhouse.
But, given the fact that companies like Google and Amazon employ people who do nothing but mimic what a speech recognition system does in an attempt to improve the accuracy I suspect you're going to have to "think bigger". As in, the minimum viable hardware is probably a Raspberry Pi (something with equivalent memory and processing).
I suggest you also consider including full acknowledgement. For example...
Human: Four tablets of gobbledygook please.
Computer: You requested four tables of gobbledygook. Is that correct?
Human: Yes.
Computer: {dispenses}
You can try a module called "Voice Recognition Module DM50A", which supports recording training and can record 50 instructions with a recognition rate of over 90%. I have bought it on AliExpress and the effect is good. You can have a try.