filtering - Time-varying band-pass filter in Python -


i trying solve problem similar 1 discussed in this post

i have broadband signal, contains component time-varying frequency. need monitor phase of component on time. able track frequency shifts (a brute force method of) peak tracking in spectrogram. need "clean up" signal around time varying peak extract hilbert phase (or, alternatively, need method of tracking phase not involve hilbert transform).

to summarize previous post: varying coefficients of fir/iir filter in time causes bad things happen (it not shift passband, confuses filter state in ways cause surprising transients). however, there some way adjust filter coefficients in time (probably jointly modifying filter coefficients , filter state in intelligent way). beyond expertise, i'd open solutions.

there 2 classes of solutions seem plausible: 1 use resonator filter (basically damped harmonic oscillator driven signal) time-varying frequency. model simple enough avoid surprising filter transients. try -- resonators have poor attenuation in stop band (if can said have stop band?). makes me nervous i'm not 100% sure how resonate filters behave.

the other suggestion use filter bank , smoothly interpolate between various band-pass filtered signals according frequency. approach seems appealing, suspect has hidden caveats. imagine linearly mixing 2 band-pass filtered signals might not expect, , might cause weird things? but, not area of expertise, if mixing on filter bank considered safe solution (one has been analyzed , published before), use it.

another potential class of solutions occurs me, take phase frequency peak in sliding short-time fourier transform (could windowed, multitaper, etc). if knows prior literature on i'd interested. related, take phase @ frequency power peak sliding complex morlet wavelet transform on band of interest.

so, guess, have 3 classes of solutions in mind. 1. resonator filters time-varying frequncy. 2. using filter bank, possibly mixing? 3. pulling phase stft or cwt, (these can considered subset of filter bank approach)

my supicion in (2,3) surprising thing happen phase time time, , in (1) may not able reject noise we'd like. it's not clear me problem has perfect solution (uncertainty principle in time-frequency resolution?).

anyway, if has solved before, and... better, if knows papers sound directly applicable here, grateful.

not sure if help, googling "monitor phase of time varying component" resulted in this: http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4124619&url=http%3a%2f%2fieeexplore.ieee.org%2fxpls%2fabs_all.jsp%3farnumber%3d4124619

hope helps.


Comments

Popular posts from this blog

c++ - Delete matches in OpenCV (Keypoints and descriptors) -

java - Could not locate OpenAL library -

sorting - opencl Bitonic sort with 64 bits keys -