What I want to do is to make an animation for the shifted impulse responses of different subbands, like this:
def get_animation(j, **kwargs):
'''
This animation shows the subband component of a reconstructed step signal
when the input signal is shifting.
Parameters:
- j: int, the subband to consider. -1 implies lowpass subband.
- **kwargs: additional arguments to pass to the DTCWT and IDTCWT.
'''
# Create a dirac signal
T = 2**10
wt = murenn.DTCWT(**kwargs)
iwt = murenn.IDTCWT(**kwargs)
fig, ax = plt.subplots()
line_ref, = ax.plot([], [], ls='--', color='gray', label="input")
line, = ax.plot([], [], lw=2, label='output')
ax.set_xlim(0, T)
ax.set_ylim(-1, 1.3)
ax.set_title(f'Level {j}')
ax.grid(ls='--')
ax.legend(loc='upper right')
def animate(tau):
# Shift the signal
x = torch.zeros(T)
x[(T//2+tau):] = 1
x = x.reshape(1, 1, T)
# Apply the transform for the subband j
lp, bps = wt(x)
if j == -1:
rec_x = iwt(lp, [bps[k]*0 for k in range(len(bps))])
else:
rec_x = iwt(lp*0, [bps[k]*0 if k != j else bps[k] for k in range(len(bps))])
# Plot the result
line_ref.set_data(range(T), x.squeeze().numpy())
line.set_data(range(T), rec_x.squeeze().numpy())
return line,
ani = FuncAnimation(fig, animate, frames=range(0,T//4,1), init_func=init, blit=True)
return ani
lostanlen
commented
1 month ago
I think it would be nice to have some demonstrations of the shift-invariant property of the dual-tree wavelet transform, as described in Chapter 4 of this paper: Complex Wavelets for Shift Invariant Analysis and Filtering of Signals
What I want to do is to make an animation for the shifted impulse responses of different subbands, like this: