Measuring reverberation time

[Jadwiga Karska]

Sorry for maybe stupid question but I could't find any answer for my problem.
I'd like to measure reverberation time of the room based on the recording with talking few people (no test signal has been done, just a recorded conversation).
Is it possible?
Any suggestions are welcome.

Thanx

 

[Wayne A. Pflughaupt]

I’m going to hazard a guess that it’s not possible to get a meaningful measurement like that.

Reverberation time (RT) is typically, and most accurately, measured using an omnidirectional signal source. Typically we want to find how long it takes for the signal to drop 60 dB to the room’s noise floor once it is shut off (RT60). In a regular room with a noise floor in the 30-40 dB range, that requires a 90-100 dB signal.

So we can see the problems with your scenario:

* Spoken voices are not omnidirectional.

* Spoken voices in conversation are nowhere near loud enough.

Regards,
Wayne

 

[Jadwiga Karska]

Thanx a lot for the answer Wayne.
Your second condition seems to be hard to achieve indeed
Thank you for the explanations

 

[Matthew J Poes]

And the means by which you typically measure RT wouldn’t work with a recording of a voice either. You would need to know what the spoken voice is like in a non-reverberant condition.

For example, if you were to try and use a recording of words to create an impulse response of said spoken words in a room, you basically do some processing of the files where the original recording of the voice (taken in a reverberant free zone) and the recording of the voices played back in the room to output what is effectively a transfer function. REW can only do this with sinesweeps. Matlab can do this with any file, but as I’ve found out, the use of non-test tones yields weird results.

Is it not possible to go into the room and measure directly?

 

[DanDan]

REW can measure Room Decay using it's Sweeps. Long and or Multiple combined sweeps can be used to improve S/N Ratio. But 75-80dB should work fine as the Sweep Method, hears down into the Noise by about 15-20dB better than using actual noise. Actual Impulses can be used also, e.g. Ballon Bursting or even a Starter Pistol. I remember an email from JohnPM saying he hope to include a feature to use music as a source also. I haven't looked in a while but that may be in place too. It might work with a recording if voices played back a bit louder than normal. But if you have a speaker in place, Sweeps.....
DD

 

[Matthew J Poes]

REW can measure Room Decay using it's Sweeps. Long and or Multiple combined sweeps can be used to improve S/N Ratio. But 75-80dB should work fine as the Sweep Method, hears down into the Noise by about 15-20dB better than using actual noise. Actual Impulses can be used also, e.g. Ballon Bursting or even a Starter Pistol. I remember an email from JohnPM saying he hope to include a feature to use music as a source also. I haven't looked in a while but that may be in place too. It might work with a recording if voices played back a bit louder than normal. But if you have a speaker in place, Sweeps.....
DD

I think he is trying to pull the RT out of an existing recording of someone talking. But I may have misunderstood.

 

[DanDan]

Aaaah, probably right. If a good loud handclap or balloon had been recorded at the time......
DD