Actually it's a pretty flat power spectrum.
Look at the total dynamic range. It's only about 10 dB. Just for comparison look at the periodogram (power spectral density) of a white noise sequence the length of your data.
A white noise sequence theoretically has a perfectly flat spectrum (PSD)
(requires signal processing toolbox for periodogram)
x = randn(500,1);
[Pxx,F] = periodogram(x,[],length(x));
plot(F,10*log10(Pxx));
Also, your signal is nonstationary. All the activity in the signal occurs very early and then the signal is essentially zero after the first few samples.
If you really want to know what the spectrum of the signal is, you should limit your analysis to just the first few samples, or use spectrogram() for a short-time Fourier transform.
