In the General case, you will need plug-ins equalizer, gate and compressor.
Especially fool makes no sense, because the gain of the digitized signal always entails the amplification of noise. And if the voice is very quiet, the result will be very noisy. So the plan is:
1. To raise the overall volume of the recording
2. To put it in the EQ those frequencies that do not transmit the sound of the voice. It's up to about 80 Hz and 10 000. Specific values depend on the nature of the voice and noise. Be careful with the frequencies above 1000 Hz, because they contain information from which we isolate the consonant.
2A (Optional). you can raise the first few harmonics of the fundamental frequency of the voice. That is, if the person is mumbling and most of the signal power of his voice is in the area of, say, 250 Hz, you can play with peaks 250*n Hz, where n is an integer. Sometimes makes it easier to understand speech. Modern equalizers often have a visualization of the frequencies, so finding peaks this is elementary looking at the screen.
3. With the help of the gate to remove the noise between speech sounds if the noise is noticeably quieter speech itself. Gate passes the signal, the volume of which is above a certain threshold. Advanced gates have flexible options that let you "get ahead" and skip the quiet signal, which obviously will soon be loud. So you can not lose the quiet beginning of a word, for example. Also advanced gates can only look on for the selected frequency range, and this really facilitates noise reduction.
4. Compressor (optional) - dynamic processing. If after the above treatments the signal is too jump in volume, you can brush it with the compressor putting compression ~4: 1. You can play with other parameters of compression, for example, with speed of response. Can also improve articulation.
Everything you need is out of the box in Reaper. First 30 days free.
Paragraphs 2 and 3 together, in principle, be replaced by dynamic equalization. In Reaper is a plugin ReaFIR in Substract mode. First let him hear a few seconds of noise from your recording, it will build a noise profile and deduct it from the original signal. Works well when at a frequency that is a voice, little noise.
But there's an easier way that may work:
There is such a wonderful product - Isotope RX - the set plagnol to restore the signals. A bunch of clever mathematics, which works wonders. The interface is elementary, the theory do not need to know.
1. Shake the trial version of RX
2. Shaking Reaper.
3. In Reaper throw his record and process plug-ins RX.