def tfmap(x, fs, fvec=(5, 250, 10, 5), fc=None, norm=0, baseline=(1, 1),
**kwargs):
if fc is None:
fc = freqvec(fvec[0], fvec[1], fvec[2], fvec[3])
# Compute the tf:
tf = power(x, fs, fc, norm=0, baseline=(1, 1), **kwargs)
# Return the normalized mean TF:
if (norm != 0) or (baseline != (1, 1)):
X = n.mean(tf, 2)
Y = n.matlib.repmat(
n.mean(X[:, baseline[0]:baseline[1]], 1), X.shape[1], 1).T
tfn = normalize(X, Y, norm=norm)
else:
tfn = n.mean(tf, 2)
return tf, tfn
def power(x, fs, fc, baseline=(1, 1), norm=0, window=None, edge=None,
width=None, step=None, split=None, **kwargs):
# -----------------------------------------------------------------------
# Check input arguments :
# -----------------------------------------------------------------------
timeL, trials = x.shape
# Number of frequencies :
if type(fc) == tuple:
fc = [fc]
nfc = len(fc)
# Split the power :
if (split is None) or (type(split) == int):
split = [split]
if len(split) != nfc:
split = split*nfc
fcSplit, fcSplitIndex = [], []
for k in range(0, nfc):
if split[k] == None:
fcSplit.append(fc[k]), fcSplitIndex.append(k)
elif type(split[k]) == int:
f = n.arange(fc[k][0], fc[k][1]+split[k], split[k])
splitList = [(f[i], f[i+1]) for i in range(0, len(f)-1)]
fcSplit = fcSplit + splitList
fcSplitIndex = fcSplitIndex + [k]*len(splitList)
# Binarize the power :
if type(window) == tuple:
window = [window]
else:
if width is not None:
if step is None:
step = round(width/2)
window = binarize(0, timeL, width, step, kind='tuple')
if (window is not None):
nWin = len(window)
# -----------------------------------------------------------------------
# Extract power :
# -----------------------------------------------------------------------
xF = n.square(brainfir.filtvec(x, fs, fcSplit, 'amplitude', **kwargs))
# -----------------------------------------------------------------------
# Remove the edge effect :
# -----------------------------------------------------------------------
if edge == 'on':
xF[:, 0, :], xF[:, -1, :] = xF[:, 1, :], xF[:, -2, :]
# -----------------------------------------------------------------------
# Normalize the power :
# -----------------------------------------------------------------------
if (norm != 0) or (baseline != (1, 1)):
xFn = normalize(xF, n.tile(n.mean(xF[:, baseline[0]:baseline[1], :], 1)
[:, n.newaxis, :], [1, xF.shape[1], 1]), norm=norm)
else:
xFn = xF
# -----------------------------------------------------------------------
# Mean the splitted matrix :
# -----------------------------------------------------------------------
xFsplit = n.zeros((nfc, x.shape[0], x.shape[1]))
for k in range(0, nfc):
if len(n.where(n.array(fcSplitIndex) == k)[0]) != 1:
xFsplit[k, :, :] = n.mean(
xFn[n.where(n.array(fcSplitIndex) == k), :, :], 1)
else:
xFsplit[k, :, :] = xFn[n.where(n.array(fcSplitIndex) == k), :, :]
# -----------------------------------------------------------------------
# Binarize the power :
# -----------------------------------------------------------------------
if window is not None:
xFsplit = n.swapaxes(xFsplit, 1, 2)
xFsplitWin = n.zeros((nfc, trials, nWin))
for k in range(0, nWin):
xFsplitWin[:, :, k] = n.mean(
xFsplit[:, :, window[k][0]:window[k][1]], 2)
xFsplit = n.swapaxes(xFsplitWin, 1, 2)
return n.squeeze(xFsplit)
