Originally Posted by joema
Fortunately there is already ready-made sophisticated audio spectral editing software for this, e.g, Izotope RX4. It easily allows a skilled forensic analyst to null out specific spectral patterns such as alarms which obscure speech.
Originally Posted by Volume
Are these statements still true for modern, digital CVR?
Yes, the audio bandpass doesn't greatly affect the ability to forensically extract speech from a cluttered audio environment. The technician visually edits audio spectrograms using software designed to identify and remove unwanted noises. It is vastly beyond tweaking an equalizer.
It is very sophisticated -- usually a cell phone ringing during speech can be removed so perfectly there is no significant artifact remaining.
The software already exists so it's not like they have to engineer something from scratch. In both music/video production and advanced audio forensics for law enforcement, audio technicians are highly trained using this software and accomplish amazing results.