signal processing - How can I select an optimal window for Short Time Fourier Transform? -

i want select optimal window stft different audio signals. signal frequency contents 10 hz 300 hz appropriate window size ? signal frequency contents 2000 hz 20000 hz,  be optimal window size ?

i know if window size 10 ms give frequency resolution of 100 hz. if frequency contents in signal lies 100 hz 20000 hz 10 ms appropriate window size ? or should go  other window size because of 20000 hz frequency content in signal ?

i know the classic "uncertainty principle" of fourier transform. can either have high resolution in time or high resolution in frequency not both @ same time. window lengths allow trade off between two.

windowed analysis designed quasi-stationary signals. quasi-stationary signals signals change on time on short period of time might considered stable.

one example of quasi-stationary signal speech. frequency components of signal change on time when position of tongue , mouth changes, on short period of time approximately 0.01s might considered stable because tongue not move fast. range of 0.01s determined our biology, can't move tongue faster that.

another example music. when touch string might consider produces more or less stable sound short period of time. 0.05 seconds. within period might consider sound stable.

there might other types of signals, example, might have frequency 10ghz , quasi-stationary of 1ms of time.

windowed analysis allows capture both stationary properties of signal , change of signal on time. here not matter sample rate signal have, frequency resolution need or main harmonics. main harmonics near 100hz or near 3000hz. important on period of time signal stationary , on can considered changing.

so speech 25ms window because speech quasi-stationary on range. music take longer windows because our fingers moving slower our mouth. need study signal decide optimal window length or need provide more information it.