Know-How Exchange : Other

It's a lot of work to clean up recordings. There are software packages out there to reduce pops, hiss, etc....but the learning curve is steep, and its time consuming to do. You could check out Acid 5, Soundforge, or similar products. You may need some VST and filtering plug ins to clean up the sound. I wonder if it may not be better to hire a transcription company to transcribe conversations, then hire voice actors in a proper sound studio to make your recordings.

Darcy Moen

Hi Carrie

I’ve only just got onto this posting and I hope that I can help to clarify the technical issues which I think you need to separate out. I hope that I understand you correctly: Here’s my summary of your requirements.

As I read it your process is to take a telephone audio feed and record it. Then you wish to incorporate that audio file into some web based content, (such as a podcast but possibly other formats). You are looking to clean up the sound by removing background noise and distortion, presumably to make it more audible, pleasant or easy to listen to, but you appear to want to do this in real time; when you make the recording, rather than treating the file afterwards.

1) I would question the need to treat the sound in real time unless you are involved in live broadcasts. Cleaning up an audio input as it happens requires a shed-load of equipment and great skill in using it either immediately or within the confines of a radio station’s 5 or 10 second broadcasting delay.

As the recording, if done to the best quality you have available, contains all of the audio information (Distortion and background noises included), selecting the appropriate treatment is easier as an after-recording process, rather than trying to get it right at the time.

After all, you are trying to take away distortion or background noise from something which is already recorded in totality. You are not adding anything to the recording, form a technical standpoint. What you are adding is subjective sound quality. In terms of the bytes you upload to your website, you are subtracting information – subtracting the information your listeners don’t want to hear!

2) Let’s look at the telephony side. If your client is recording telephone to telephone interviews, then both the interviewer and the interviewee will be subjected to the frequency and dynamic limitations of your telecoms system. This is important with regard to treating a resulting audio file because the data sources are both subject to the same technical constraints.

If the interviewer is studio based, their audio feed will be of much higher quality both in terms of frequency response, dynamic range and probably background noise. In addition the quality of their microphone will be a lot better. This can result in the “Telephone” end of the interview sounding poor and tinny in comparison to the studio voice, which, psychologically, renders it harder to comprehend by listeners.

3) Here’s some limitations placed on the sound quality you get from a telephone feed – they affect everything that you can do with the caller’s audio input.

Telephony bandwidth is limited, the signal is compressed into a narrow dynamic range, the specification of caller’s microphones will not always be up to standard and of course, depending on their physical location, the background noise will be variable.

International standards vary but rarely exceed 200 – 5000Hz (300 – 3000Hz is one standard) and a dynamic signal to noise ratio of 42-48dB. VOIP standards vary depending on modem speeds and can vastly exceed the above if used in conjunction with high quality conferencing telephone headsets. Unless you issue your callers with a standard headset and a soundproof booth (!) there’s no way round this. Clip-on microphones are a bit of a red herring at the caller’s end.

You also refer to Distortion and Background Noise as if they were the same thing. It is important to differentiate between the terms because you treat them differently.

4) Distortion is any change to the original audio sound and it is produced by the telephone microphone, the telephone line, the electronics at your end and the conversion process of taking an analogue signal and turning it into a digitally stored file. There are also distortions associated with VIOP telephony such as aliasing.

Certain types of distortion can be compensated for easily (Such as a telephone microphone with a poor frequency response) and some (Such as clipping, where the microphone produces a grating sound or the pre-amplifier achieves the same sort of effect) are much more difficult to sort out, requiring sophisticated audio enhancement software and / or hardware to extract the audio information.

5) Background noise is more of a problem and your ability to remove it will depend on its consistence. Noise cancellation software is widely available and used by the military and the police and more recently by Bose Inc. It requires a degree of consistency in the background noise, such as that from a helicopter, heavy traffic, in-car road and engine noise and noise in a commercial airliner. It works by recording the pseudo-constant background noise over a short sample period and then electronically subtracting it from the constantly varying audio signal.

Leaders in this field are NCT with ClearSpeech:

https://www.nctclearspeech.com/nctabout.htm

Who provide background noise and echo cancellation for telephony, in car systems and formula 1 motor racing. There’s also a wide range of noise cancellation products discussed on:

https://www.electronicstalk.com/scripts/?FAC=search&prefilled

Where you will find some references to your distortion reduction problems – i.e. filtering the audio feed or digitally enhancing the voice component to the detriment of the distortion.

6) Distortion is a different matter, requiring hardware or software processing of the signal contained in the audio recording. Some techniques here can remove background noise as well, such as mains hum or other totally constant background interference. Clicks and plops can also be taken out, but clipping distortion from a bad microphone is notoriously harder to eliminate, requiring a complex extraction of the underlying voice information and the removal of the clipping harmonics. Rather than go into 5 pages of detail here, there’s an excellent article on:

https://preview.videosystems.com/news/video_salvaging_bad_audio/

This covers all the bases and has numerous references to hardware and software to assist you.

7) Your AudioAcrobat software. This is a package which facilitates recording and posting audio files. It offers little in the way of signal-treatment and as you say has limited ability to work with add-ons, but why should it – it’s not its function.

If you are happy with the way it posts and manages your audio to web transactions, then the best solution is to treat the raw recordings separately to produce a file which will be posted by AudioAcrobat. After all, it doesn’t care where the audio file comes from and choosing to record a file, manipulate and treat it and then post it via AudioAcrobat would at least allow you to continue to use one element of your existing system without the upheaval needed to change it.

In summary follow this schema:

• Record your interview
• Treat the audio file for distortion via hardware and / or software
• Treat it for background noise according to the nature of the noise by software
• Post it to your website

I’m sure that there are software and hardware combinations which will do the lot for you – one of my clients runs a video and recording studio and produces podcasts and streamed videos, but the investment is substantial and he’s looking to adhere to studio quality across the board. To him, a telephone feed is crap audio, but it’s crap audio which is very easy to treat because he only has to concentrate on annoying things in that very narrow bandwidth

Hope that this helps – I know that it’s not a roadmap to a total solution, but we only have limited space here. You could get a freelance audio engineer to recommend a solution for you to match your exact requirements – I’d say that it would take a day of consultancy and perhaps a couple of days to set everything up and train your client into using it. It might cost you $500/ day.

Regards


Steve Alker
Unimax Solutions