Removing Reverb and Resonances

August 20, 2020

Share


In an ideal world all recordings would be done in an optimal environment which exactly features the desired characteristics (reverb length, early reflections, frequency response ...). Unfortunately, in many cases such an environment simply does not exist or is too difficult to come up with. To acoustically optimize a recording room is a complicated and oftentimes very expensive process which requires a lot of expert knowledge. Additionally, if you are doing any live recordings it is usually simply impossible to choose or change the setting. This article will outline the major problems of sub-optimal recording environments and some measures that can be taken to tackle the negative effects. We will focus on the two major problems which are resonance frequencies and unwanted reverb.


Resonance Frequencies

A resonance frequency corresponds to a specific wavelength which gets amplified due to the geometry of the room. If the wavelength or a multiple of it exactly fits in a room between to parallel surfaces a so-called standing-wave will occur. For example a 200hz sinus has a wavelength of 1.72 meters. If your recording room happens to have exactly a width of 1.72 meters (or also 3.74 meters) and the walls are plain, reflective surfaces you will observe a boost of this frequency in your recordings (and possible also of other harmonics as 100hz or 400hz). The physical measures to get rid of this problem is usually to modify your walls to have fewer reflections or more diffuse reflections by installing specific room acoustic elements. If the room cannot be changed or the recording is already done, it is possible to apply a static EQ to reduce the unwanted resonances:
Resonance EQ
Here, the obvious challenge is to identify the frequencies by ear which requires some amount of practice but is definitely doable. By using any parametric EQ and sweeping across the frequency spectrum the room resonances will become obvious and can be reduced as much as necessary.


Reverb

The second effect which can become a problem is room reverb. In every recording environment you can think of there will be some degree of reverb (except maybe if it was done in an unechoic chamber). Without any reverb, the recorded source would sound very unnatural since our ears are not used to listening to only the direct sound without any room reflections. However, if you are recording for example a lecture in a large lecture hall, the reverb might be quite annoying and definitely doesn't benefit the speech intelligibility. In contrast to room resonances, here, not just single frequencies but the entire spectrum is affected. A static parametric EQ is most likely not going to fix this issue and other tools need to be applied. One type of plugin which is worth a try is an expander:
De-reverb Expander
An expander will amplify all input levels above a threshold while levels below that threshold pass unaffected. The idea is to place the threshold exactly below the lowest level of the input speech so that only the speech gets amplified and the reverb tail doesn't. This can improve the signal to reverb ratio and the overall signal will sound a bit cleaner. If you have a multiband expander at hand you can even tune this process for different frequencies to handle the spectrum of the signal more accurately. However, this approach may have some undesired side effects since it also changes the overall speech dynamics.


Accentize DeRoom

Our DeRoom plugins can simplify the before mentioned processes. Using a machine learning algorithm the resonance frequencies will be automatically detected and suppressed. This process will be adapted in realtime. So even if the talking person is moving and changes location the resonances will be re-calculated without requiring any further parameterization by the user. Also the reverb content of the signal will be estimated and filtered out by an intelligent time-varying filter which doesn't affect the speech dynamics as we saw in the case of using an expander. We offer a standard version of DeRoom and also one for professional contexts: DeRoom Pro. The following table outlines the difference between both versions:
DeRoom DeRoom Pro
Room Resonance Suppression
Reverb Removal
Different Room Sizes
Automatic Room Size Detection
Spectral Focus Mode
Attack/Release Controls
Only Reverb Listening
Low Latency Mode
Spectrogram and Gain Visualization

DeRoom will be sufficient if you for example record your own podcast or youtube video in your living room and want to reduce the reverb to make the listening experience more pleasant. However, if you do professional post-production work especially the spectral focus mode of DeRoom Pro which allows custom multiband processing will come in very handy to fine tune the results. Also the difference listening mode will help you to make sure you are not taking away too much speech signal components.

DeRoomPro