A Visualization Scheme for Complex Spectrograms with Complete Amplitude and Phase Information

Abstract

We present a novel visualization scheme for complex spectrograms of speech and other non-stationary signals, in which both the amplitude and phase information is visualized simultaneously using the conventional RGB (red, green, blue) color model. The three-layer image matrix is constructed in such a way that the absolute values of the real and imaginary parts of the time-frequency analysis of speech are used to fill the first (red) and third (blue) layers, respectively, and an encoding scheme is adopted to represent both signs of the real and imaginary parts and the computed values based on this encoding scheme are used to fill the second (green) layer. The importance of phase in spectrogram applications has been gradually recognized, and imaging processing techniques are increasingly used in spectral analysis of non-stationary one-dimensional (1D) signals such as speech signal. The ability to present complete spectrogram information (both the amplitude and phase information) on a single RGB image is potentially useful, especially for those applications in which phase information plays a complementary role.