Human Time-Frequency Acuity Beats the Fourier Uncertainty Principle
arxiv.org
arxiv.org
Presumably, what they found is that human's auditory system is not doing simple operation like that but has the ability to pick out frequencies in short time segments (IE, while listening in real time) with better resolution that a short-time fourier transform. IE, somebody could estimate the frequency of a pitch faster than what you'd expect by the equation I gave above.
None of this violates the laws of physics; the ear and brain would just need to do a different type of analysis. Probably more like wavelets.
Is there a possibility that overlapping DFT windows could get you increased accuracy? I mean, instead of looking at each 0.3s individually, you could look at 0.3s segments, starting each segment 0.1s after the previous segment started so that there is some overlap.