Voice analysis, audio clarification and audio evidence cleanup in a way that is admissible in court

One of the main purposes of Forensic Audio is to improve listenability or the intelligibility of the audio.  Here's some common things that are done in forensics:

Listenability enhancement is a relatively simple procedure.  You use equalization, compression, and various computer programs that will increase amplitude (loudness), control distortion and/or remove background sound from an audio recording.  Usually the audio is already intelligible, but needs to be made more listenable.

Intelligibility enhancement is complex and demanding.  This can involve disassembling the audio track into smaller sections, processing each differently and then reassembling it back into a mirror image of the original.  The person doing this must have a deep understanding of sound physics, human auditory perception, and phonetics as well as extensive experience in the procedures of professional audio.  The end result can be a significant improvement over the original material, with previously unintelligible dialogue becoming intelligible and, in some cases, previously unknown audio events being heard.

Authenticity analysis is used to determine if a recording actually happened at the same place as the acoustic events it represents.  Suspicious recordings may have their authenticity verified or negated by physical inspection, aural analysis, tape manufacture date verification, frequency, energy, amplitude and formant structure analysis, magnetic tape development, magnified visual inspection and photographic documentation of the results.  Whew.

Event sequence analysis is the analysis of the sequence of audio events to determine such things as timing, frequency and amplitude.  This was a technique used to determine the timing between gun shots for the JFK assassination investigation.  This can also verify or negate the authenticity of a recording.  If, for example, on the same tape, two separate events are exactly the same regarding timing, amplitude, dialogue, frequencies, and so forth, then it is a statistical certainty that the recording has been tampered with.  The probability of a naturally occurring exact duplication of sequential audio events is infinitesimal.

Other signal analysis is used to evaluate events which have been recorded but which may seem extraneous to the investigation.  Such audio events may reveal important information, such as background thumps or conversation, machine rhythms, engine frequency, reverberation, or even gunshot discharge signatures.  Computer analysis and screen display/comparison of these other signals can reveal information that can be of extreme importance when forming opinions and/or determining causes or circumstances of recorded events

Success requires work on material as close to the original, if not the original, as possible.  A perfect playback of a copy of a badly aligned original tape will sound as bad as the original.  Sometimes it may even be best to use the original recorder if it is available.

The person doing this analysis must maintain proper chain-of-custody of the material as well has a detailed record of what was done, why, and the expected end result.