Page 1 of 1

how to remove echo from a speech?

Posted: Thu Jun 08, 2006 7:41 am
by ahmad
Thers too much echo in a record coz it was a recorded in an empty lecture theatre, how to remove it? o_O
Thank you

Posted: Thu Jun 08, 2006 4:00 pm
by DougDbug
:( As far as I know, it's technically impossible.* Some things can't be reversed... How do I un-fry an egg, or un-bake bread? There are really good reasons that recording studios are soundproof, and have sound-absorbing materials on the walls. (It is easy to add artificial echo and reverb.)

You might be able to make some improvement by using the Equalizer to remove any low-frequency "boom", and some Noise Reduction and / or Noise Gate might help a bit too.



* It might actually be possible with a super computer and the right software. :? But, I don't think there's any commercial software available... maybe something in a university lab somewhere... :?

Posted: Thu Jun 08, 2006 4:05 pm
by Togglehead
^^^what he said

Posted: Thu Jun 08, 2006 9:55 pm
by piano nick
I would agree with the others.

If one considers what echo or reverberation is, it may be easier to understand:

Echo is severe or heavy reverberation, or in other words, reverberation taken to extreme is echo. Both are caused by the difference it takes for sound to travel from the source to your ears (or to the microphone).

The shortest distance from the sound source to your ears is a straight line, however in a large hall with hard (reflective) walls, the sound also travels in multiple paths to the different walls, then is bounced to your ears. These other travel paths are longer and the sound arrives at your ears a bit later in time, so in reality you are hearing the original sound many times over, each one separated by slight differences in time (there are also some other effects that are a function of the absorption of the walls).

With very short differences in travel time, we call it reverberation; longer time differences are called echo.

The problem is that it is the same set (or similar set) of sounds, arrive at your ears (or the microphone) at different times. How can you separate them out? Which ones do you remove? The most powerful computer known to man that can sort out the sounds is the human ear/brain connection. Personally I doubt that non-human computers will ever do it.

I always recommend (as Doug has alluded to) that when recording, record flat with no reverb - it can be added very nicely during processing, but it can't be removed without some considerable loss is quality.

This problem is really the same as separating different sounds from the wave file. A frequent question on this forum is "how do I remove xxxxx from my recording", where xxxxx is (swearing, whistling, singing, drums, yelling, echo, reverb, etc). Sadly the short answer is "record again".


PN

Posted: Fri Jun 09, 2006 1:05 am
by donrandall
You have been getting some good advice from some guys that know their stuff - one thing that was suggested may work, even if other attempts gain you nothing.

It will require some trial and error. Make a copy of the file - maybe just a minute or so, that should be enough to experiment with.

Try using the graphic equalizer to lower (attenuate) a given frequency range while listening to the file using the "preview" button on the equalizer drop down box - it will be a little green arrowhead on the right side of the box - try different bands until something becomes apparent. By that, I mean that it should soon become clear whether attenuating (lowering the volume of) a frequency range will reduce the echo or not. If you find that it is helpful, you may get all the result you need with the graphic eq - or you might want to use that information to switch to the parametric equalizer, which will give you more control.

Posted: Fri Jun 09, 2006 3:42 am
by piano nick
Ahmad:

Ask Don more questions - he works with voice whereas I work with instrumental music (mainly piano).

There may be others that can offer some advice on lessening the effects of excessive reverb (echo).


PN

Posted: Fri Jun 09, 2006 12:01 pm
by Togglehead
Im an audio engineer for a teleconference company. I handle all the digital and analog media that enters and leaves this building. Im also responisible for making sure the telephone interfacing circuits are in the right shape...and work. In other words, if something goes wrong with any audio in this office....i have to fix it. heh heh

There are SO many things to interfere with my recordings over telephone lines. One of the biggest differences about this field, than a typical recording job, is that i have almost NO control over the recording input. Its a phone line, i work with what they give me, and what the other end has for equipment.

When it comes to mono voice and distortion, ive seen it all. And let me just tell you, notch filters are your friend. First of all, with voice, EVERYTHING is different. For example if you try to use the clipboard option in the noise reduction filter, it will most likely ONLY be effective for THAT word. The timbre of everyones voice is very different...and most often, global file filters dont work. It gets pretty tedious.

Also, remember peoples breath. If people dont breathe, they cant speak. This is one of the reasons that edited audio stands out so much....people dont leave breaths and inflections in the right place. Use the trailing end of words, and other edits, to sort of "complete" the edits you make, by making them sound more natural. With voice, there is only one thing to focus on, and if your edits arent clean, then that is exactly what will stand out.

Something ive never tried, but thought about:

1.Copy the audio track into two identical mono tracks (if they arent already)....do what don said, and use like a minute or less to keep sampling time down.

2.Merge tracks into a stereo file, one left, and one right. This is basically called dual mono, since there is no stereo separation.

3."Bump" one audiotrack forward by a few milliseconds...chances are this will sound worse. Do it again, and again, and again, until MAYBE...just MAYBE some of the echo cancels out. You might have to lower the volume of the bumped track to not get confusing.

Its just a thought, since harmonics can do some interesting things....lol

Posted: Fri Jun 09, 2006 5:00 pm
by Blandine Catastrophe
To remove like you said, try to use the clipboard/mix, by adjusting the value of the delay, and of the level. For the delay adjustment, you must re-launch the play button of the effect monitoring.

When approaching the good value, you may see some stripe in the spectrogram, that can give you an idea of the fine adjustment as the frequency difference between two stripes corresponds to the period.

Possibly, another fact can add some difficulties, the spectral colour of the echo can be different. Some phase differences can be found between different frequencies. That should let a lot of remaining parts of the echo, and possibly you can only reduce it, instead of remove completely.

Re: how to remove echo from a speech?

Posted: Tue Jun 08, 2010 2:32 pm
by wfala
Come on you guys... Not that trauma. Everything is possible when you're talking about GoldWave...
I needed the Echo reduce and i did it, not perfect that's true. But i reduced it 80% at least, which is a great resault.
What to do:
1- Go to the ECHO effect and do a STEREO REVERB preset. It will send a feedback to the file simulating a different enviroment to the record.
2- Go to the Compressor/Expander menu and do a NOISE GATE 2 or 3 preset, which eliminates the bad consequences of the reverb.
3- Then you go to the Pan menu and try to identify which one sounds better and clear, if it's the right or the left side of the audio chord. Then you chose the better one and we'll duplicate it soon. Pick the better side.
4- Got to the Edit menu and run down to the Channel item, then pick the side you selected earlier and copy it to clipboard. Then select the other channel and paste it there, duplicating the better channel.
5- After that you should go to the Equalizer effect and equalize each channel differently. You can set a boost bass to the right and then boost the treble to the other. But that's only a sugestion.

After all that your record should sound more clear.
Thank's

Re: how to remove echo from a speech?

Posted: Mon Apr 16, 2018 11:47 am
by titania
Sorry to drag up an old thread but it is possible nowadays to remove reverb from recordings. If you search YouTube for "Remove Echo and Reverb from Dialogue" there is a review of two similar "deverb" VST plugins. These plugins are clever enough to cancel out the reverb that has been recorded.

I guess the software figures out the impulse response for the reverb and then applies its opposite.