Effect - Equalizer allows you to adjust selected frequency bands. It’s usually easier to start with the Graphic EQ option.
Try pulling-down all of the sliders below 100Hz to see if that moves things in the right direction. If that leaves too much mid-bass, try reducing the bands up to 150 or 200 Hz… Just experiment to see what you can do. That should leave you with too little bass. From there, you can push the sliders back-up one-by-one 'till you find a good bass-balance. (Use the Preview button to find the right sound before “permanently” applying the effects.)
You may find that there are only one or two frequency-bands that are over-resonant, and the other sliders may need little or no adjustment.
After applying effects that can affect volume, it’s almost always a good idea to normalize the volume so the peaks are at (or near) 0dB which is the “digital maximum”. The Amplify effect will scan your file and automatically default to whatever adjustment is needed (up or down) for 0dB peaks.
Audacity itself uses floating point so it can go over 0dB temporarily/internally. But normal WAV files and CDs will clip (distort) if you go over 0dB and your digital-to-analog converter is also limited to 0dB.
There’s a lot of overlap, especially when you include harmonics & overtones.
One common trick the pros use is to set-up a parametric equalizer with a notch (a deep cut over a fairly narrow frequency band) and then “sweep” across the frequency range to find “problem areas”. (I don’t know if you can do that with Audacity’s built-in effects, or if you’d need a 3rd-party plug-in.)
The best option would be to have a real-time parametric peaking filter, where a slider would sweep the center frequency. Bandwidth and gain had to be adjusted separately.
I use a Nyquist workaround.
My plug-in boosts the sweeping center frequency by about 10 dB (it is easier to make out resonances that way). The audio is being played and the center moves upwards.
You simply press cancel as soon as you hear the boominess and the reversed filter is applied e.g. cut by -6 dB.
The breakpoint can only be determined within 1 s (at 44.1 kHz), thus the audio has to be listened too for quite a while if the step-size (in Hz) is small.
The following code in the nyquist prompt shows what I mean: