Audio Restoration vs. Enhancement Overview

[Craig Maier]

Folks,

Attached, you will find an Overview of Audio Restoration vs. Audio Enhancement which may be of interest to you.

[Brian]

Hi Craig
A very clear overview. Would it be useful to include a few sentences on gain normalization?
Thanks again
Brian

[Craig Maier]

Good idea. I will go back and take a look at that in the next day or so.

[Audyossey]

Attached, you will find an Overview of Audio Restoration vs. Audio Enhancement which may be of interest to you.

Good overview. In my experience as an amateur, the line between "restoration" and "enhancement" is a thin one. I find it's often impossible to remove noise without also removing some musical content, usually high end material, and you need to do something to "restore" that loss. And that's necessarily subjective since you don't have a pristine reference to compare it with.

There's a psychoacoustic phenomenon involved here, especially with 78s, because the ear gets confused by hiss and other continuous noise in conjunction with music. The noise makes the music seem brighter than it actually is, which is why so many people shut off the Dolby circuit in cassette recorders. The clean recording sounds dull to them. Hence, pure "restoration" is almost never a satisfactory objective.

I'm also finding that it's important to listen to your work at a consistent loudness level. I usually normalize files at -5 or -6 dB to provide headroom before fiddling with the equalization curves. I also am starting to do most denoising on the flat recording. But, if I don't normalize the recording or raise the playback volume after applying the proper equalization, I find it's easy to misjudge the bass content of the recording. Just a few nights ago I neglected to reset levels and added a bass boost at the same time I was adding a little equalization to the top end. When I finally normalized, it sounded like a musical earthquake. My computer desk practically fell apart.

It's never as easy as a simple overview makes it seem, but it's always a challenge and always fun---and addictive.

HB

[Craig Maier]

HB -

Those are all good and accurate observations. Of particular interest is the psyco acoustic phenemonen that you described invovling hiss. I have demonstrated this principle in a very simple manner. I have taken a high quality CD recording and ripped it to Diamond Cut. Then, I have used the make waves feature to create random noise. Then, I have used the Paste Add feature to add the noise to the recording at a relatively low level (-45 dB) and then played the before and after for friends and relatives. Often, folks say that the recording with the hiss has more highs despite the hiss, although when queried, they would prefer the hiss to be gone. Interesting.

Craig

[Geebster]

I find it's easy to misjudge the bass content of the recording.

How true! And it's not just us amateur's that have the problem. Many "so called" professional sound engineers have the same problem. For example, if you ever care to listen to RUSH's "Vapor Trails" CD, it is about the worst attempt of a professional engineering job that I have heard in recent years. The CD, musically, is outstanding, but the recording job done on it is just plain lousy. The recording is almost completely oversaturated with bass from start to finish. It is so bad that in certain sections you can even hear digital dropouts.

It's funny because their (RUSH) "Caress of Steel" CD (album), has the EXACT opposite problem... weak bass throughout. It always amazes me how every single person "hears" music differently. I guess that is what makes restoring music so much fun and challenging at the same time.

Regards,
GB..

[Craig Maier]

Geebster,

If they are producing digital dropout due to the bass, that means that they are completely exceeding the dynamic range of the system. This is a form of Intermodulation Distortion - - - the type of distortion which sounds the worse of all the various categories! It is kind of amazing that they would have done that - - - clearly not an act of artistic expression but plain and simply - - - an error.

[Geebster]

If they are producing digital dropout due to the bass, that means that they are completely exceeding the dynamic range of the system.

Craig:

Don't I know it! I am by no means on a level with producers and engineers that Groups on the caliber of Rush use (and I am deaf in one ear!!), but I could have done a better job on that CD in my sleep!!

It's funny, but almost every single review of that album (Vapor Trails), comment on the audio problems with that release, which really is a shame, because it is a very strong musical release.

GB..

[Audyossey]

Of particular interest is the psyco acoustic phenemonen that you described...

I remember Dave Ranada talking about the nature of hearing at a meeting of the Boston Audio Society in the early '80s. The ear works the same way as our eyes do (or our sense of touch or any other sense for that matter). A digital recording may take 44.1-, 48-, 96- or whatever thousands of samples of "sound" per second, but our ears operate at 16 Hz. Just 16 samples per second! But unlike a digital device which records only the sound pressure as a number at the sampling point, each ear sample is a complete waveform built up within that sixteenth of a second. I suppose it's analogous to what you would see in a waveform analyzer which was set to show samples in .0625-second chunks. Technically, the waveform within the ear sample consists of thousands of data points. The hair-like nerves along the cochlea are each sensitive to different frequencies and accumulate charges proportional to the presence of their respective frequencies in the spectrum of sound. The ear sample is a composite of all these records.

So in the visual world you have the well-known persistence of vision. In the auditory world you have "persistence of hearing," if you will. Visually a light will be seen as flashing on and off if it's pulsing at less than 16 Hz, flickering at around 16 Hz, and as constantly on if it's well above that rate (hence movies and television). In audio, the comparable phenomena for repeating audio events are clicks, echo, and reverberation.

Since our senses sample sensations at this relatively slow rate, our brains do a lot of educated guess work to extrapolate the missing information which helps account for optical (and audio) illusions. (This is also why the things people with "golden ears" hear and the rest of us don't can't be confirmed in properly controlled listening tests.) And, I suppose, it's why our ears can't accurately separate hiss from the high frequency content of music.

HB

[Craig Maier]

HB -

Put another way, our sense of hearing operates in the frequency domain rather than the time domain like a microphone. So, the frame rate could be looked at as occuring at 62.5 mSec intervals with the frequency resolution being determined by the number of hairs in the cochlea. Essentially, this is the process of conversion from the time domain to the frequency domain much in the same manner as a Fourier Transform (actually, a Fast Fourier Transform or FFT) is performed in the Diamond Cut Software in some of its algorithms. The ear is more than a "microphone" style transducer, but it also converts the audio signal into thousands of parallel paths which are sent to the brain. Much of the signal decoding has already been done before these signals are perceived by the brain because of this process. As you point out, vision works similarly with the optic nerve carrying thousands of parallel signals to the brain.

[DougMac]

Interesting Stuff

Seems there's more to seeing and hearing than physics. I recently saw a story about a young man injured in an accident. He could not recognize his mother or father as such by sight. If his dad stepped out of the house and called him on his cell phone, he immediately recognized him.

It turns out that there is a path between the "recognition" part of the brain and the "response" part. This was broken by his injury. He could recognize his mom or dad, but he had no emotional response. The only way his brain could resolve the dichotomy was by assuming it wasn't really his mom or dad. The good news is that over time, his brain rewired to the point he could finally realize who he was seeing.

As far as hearing goes, I think I'm losing hair both on top of my head and inside my head!
Doug

[Audyossey]

...our sense of hearing operates in the frequency domain rather than the time domain like a microphone.... So, it also converts the audio signal into thousands of parallel paths which are sent to the brain....

The joy of being an engineer and actually knowing what you're talking about! Wonderfully concise. And understandable even to someone like me who has trouble with basic algebra. Thanks for the summary.

HB