Minor improvements to cross_morlet.

from scipy import signal

def morlet(x, freqs, sample_rate, win_len=4, ncycles=5, ret_basis=False,

           ret_mode='power', normalise='wikipedia'):

def morlet(x, freqs, sample_rate, win_len=4, ncycles=5, ret_basis=False, ret_mode='power', normalise='wikipedia'):

    """Compute a morlet wavelet time-frequency transform on a univariate dataset.

    Parameters

        Array containing morlet cross wavelet transformed data [nfreqs x nsamples x nchannels x nchannels].

    """

    if ret_mode not in ['power', 'amplitude', 'complex']:

        raise ValueError("'ret_mode not recognised, please use one of {'power','amplitude','complex'}")

    # Run standard wavelet decomposition return complex values

    wt = morlet(x, freqs, sample_rate, win_len=win_len, ncycles=ncycles,

                ret_mode='complex', normalise=normalise)

    wt = morlet(x, freqs, sample_rate, win_len=win_len, ncycles=ncycles, ret_mode='complex', normalise=normalise)

    # Preallocate output array

    # [nchannels x nchannels x nfreqs x ntimes]

    S = np.zeros((wt.shape[0], wt.shape[0], wt.shape[1], wt.shape[2]), dtype=complex)

    # Preallocate output array [nchannels x nchannels x nfreqs x ntimes]

    S = np.empty((wt.shape[0], wt.shape[0], wt.shape[1], wt.shape[2]), dtype=complex)

    # Main loop

    for ii in range(wt.shape[1]):

        Flag indicating which normalisation factor to apply to the wavelet

        basis (default is 'wikipedia') - can be one of:

        * None - no normalisatio is applied

        * None - no normalisation is applied

        * 'simple' - wavelet is normalised by its own sum