Sensor for detecting particular human talking.

[amitcsedu]

I am looking for a sensor which can sense the talking behavior of a person.
for example it can be set in the particular human body. When he talks the sensor will work. If anyone beside him talk the sensor will not work. It can be voice cord base sensor.

Let give a example.
There is a room. There are 5/6 man are talking altogether. A,B,C,D,E are the person of the room.
Now i want to see whether "person B" is talking or not. The sensor will sense when particularly "person B" is talking. If Person A or C or D or E will talk the sensor will not sense.


Can any one give the sensor for doing this task.

I will be thankful to you.

With Regards,
Amit Kumar Das.

 

[dougy83]

This problem is called the "cocktail party problem/effect" and may call for algorithms to perform source signal separation. This would call for one or more microphones feeding into a digital signal processor.

Have a look at blind source separation: https://en.wikipedia.org/wiki/Blind_signal_separation - this page leads to others describing some methods to use (e.g. ICA).

If you have a multiple microphones, it is easier to separate the sources (speakers). Something less intensive would perhaps be to simply triangulate the source location by using the signal delay times to each of four microphones.

 

[amitcsedu]

Hello Dougy83,
Thanks for your reply.
Can you please refer me a sensor for doing this task?

I will be thankful to you. Is there any build-in sensor in the market?

With Thanks and Regards,
Amit

 

[misterT]

If you google "throat microphone" you will get some results. Like this one: **broken link removed**
Is that something you are looking for?

 

[dougy83]

If you google "throat microphone" you will get some results. Like this one: **broken link removed**

That would allow the simplest, most robust and reliable scenario for detecting when person B is speaking (as it excludes everyone else's voice)... therefore it likely won't be applicable.

 

[amitcsedu]

Actually the concept is like that.
Let,I want to observe "Person B". "If Person B" is talking only then the sensor will send some data in a server.
Is there any way to do this?
Actually i want to find specifically a particular man is talking (over phone/normally) or not. If a particular man is talking then the sensor will sense this and give some data to the server. The idea is like that.
I don't want some headphone to talk with someone. I just want to detect.

Can you please suggest me this type of thing?

With Regards,
Amit

 

[misterT]

Why throat mic is not suitable? You connect the output of the mic to a detection circuit which sends the data to your server. HINT: many of the throat microphones come with a radio. You can use that to send the sensor output to your server room, where you have all the detection circuitry and dataloggers. You can even find some with bluetooth, so you can pair multiple of them to a single PC.. that is the easiest solution you can come up with.

PS. I hope this is for some kind of academic research, not to monitor phone salesmen at work.

 

[atferrari]

Maybe he wants to spy Person B...

 

[jpanhalt]

I am not at all clear what the limits are. For example, the OP says,

I am looking for a sensor which can sense the talking behavior of a person.
for example it can be set in the particular human body. When he talks the sensor will work. If anyone beside him talk the sensor will not work. It can be voice cord base sensor.

Both of those underlined comments seem to imply the person being observed is aware he is being observed. More recently, it appears that will not be the case.

BTW, I don't think any implant in the vocal cords would be a good idea.

John

 

[audioguru]

Does anybody make computer software that detects a certain voice not just the words?
Does the software need "training" on millions of common words?
Does the software still work when the person has a cold or is hoarse from cheering?

 

[dougy83]

Does anybody make computer software that detects a certain voice not just the words?
Does the software need "training" on millions of common words?
Does the software still work when the person has a cold or is hoarse from cheering?

1) Yes. 2) It will need training, but not on millions of words. 3) it's unlikely due to the obvious difference in voice, but could perhaps use a synthesised 'cold/flu' filter to estimate the frequency distortion (or train another model where the person is hoarse - but this is inconvenient).

Here's an example voice (i.e. speaker) recognition project https://code.google.com/p/voiceid/

 

[amitcsedu]

Maybe he wants to spy Person B...

Hello atferrari,
I don't want to spy "Person B". I just want to distinguish all the voices particularly. That is the point.

 

[atferrari]

Hello atferrari,
I don't want to spy "Person B". I just want to distinguish all the voices particularly. That is the point.

OK. You have this very expensive **broken link removed**.

Interesting, isn't it?