Ask Your Question

Spectral density of a signal

asked 2012-07-17 09:48:47 +0200

v_2e gravatar image

updated 2012-08-02 07:31:56 +0200


Could somebody please help me find a way to build the spectral density function for a given signal in Sage?

Thank you!

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2012-08-02 07:11:13 +0200

v_2e gravatar image

updated 2015-07-03 21:08:35 +0200

A self-answer:

The power spectral density (PSD) may be defined as

$ S(\omega) = \lim \limits_{T \to +\infty} \frac{\left \vert F_T(\omega) \right \vert ^2}{T} $,

where $ F_T (\omega)$ is the Fourier transform defined as follows:

$ {F}_T(\omega) = \int \limits_0^T f(t) \exp(-i\omega t) ~ dt$

def PSD(time_series):
    import scipy

    signal_length = n(len(time_series)*(time_series[1][0]-time_series[0][0]))
    signal_fft = scipy.fft(zip(*time_series)[1])

    spectrum = []
    for i in range(len(signal_fft)//2):
    return spectrum

The accepted data set format is:

data = [(t1,y1),(t2,y2),...,(tn,yn)]



for such signal will return the Power Spectral Density of a signal.

Sometimes it is useful to apply some kind of window function to a signal prior to calculating the PSD, since the sharp start and end of the data record may produce some spurious spectral components.

Here is an example of the popular Hanning Window application for the time series:

def hanning_window(time_series):
    ''' Applies Hanning window to the time series.
    Accepted data format is a list of tuples [(x1,y1),(x2,y2),...]'''

    series_length = n(len(time_series))
    for i in range(len(time_series)):
        processed_signal.append((time_series[i][0], n(time_series[i][1] * \
                                 (0.5*(1 - cos(2*pi*i/(series_length-1)))))))

    return processed_signal

The result of its application looks like this: The result of Hanning window function influcence on the signal

One can simply call


to get the power spectral density for a data set with Hanning window function applied.

You can compare the results of spectral density calculation for the initial time series and "windowed" time series: image description

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools


Asked: 2012-07-17 09:48:47 +0200

Seen: 1,036 times

Last updated: Jul 03 '15