I am currently working on a dynamic equalizer to improve a synthesized voice's intelligibility in noisy environments.

The current implementation of my algorithm simply applies a parametric EQ to the voice at the dominant frequencies of the ambient noise (ex : if a peak around 180Hz is detected in the acoustic scene - from a fan for example - the algorithm applies a PEQ at 180Hz).

This implementation does not take into the account auditory masking phenomenons. It seems to me that it would be more clever to boost the voice at neighboring frequencies in order to try and make it so the voice becomes the masker.

Is it really better to try and tackle the problem this way ? If so, what resources can I look at in order to get some inspiration on how to implement this algorithm.

  • $\begingroup$ You do have to be more specific but some possibilities are: 1) Use auditory research to move the voice band to math human female range; there are studies that female voices are more intelligible. 2) You seem to be tending towards "shouting over" interference. Probably not a good idea. 3) Unless you are really really tight on money; move the voice to a frequency area where the incoming noise isn't. Humans are amazingly capable of sorting through incoming signals and receiving information based upon the tone. Think of a cocktail party or a bar. $\endgroup$ – rrogers Jun 11 '19 at 20:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.