Time-Frequency Analysis Using Short Time Fourier Transform

Abstract

A time-frequency analysis, which represents the time change of a signal, is significant in all fields. Wigner distribution, a short time Fourier transform (STFT), a kernel method, and a characteristic function method etc. are known as a time-frequency analyzing method, and those methods have merits and demerits. STFT method has been used as ultrasound blood-flow imaging for a long time, because it is suitable for a non-stationary signal analysis. In this paper, I investigated how to control a time-frequency resolution of STFT, and evaluated the image quality of non-stationary signals using a point-spread function (PSF). The time-frequency resolution of an image corresponds to the aspect ratio of a pixel. Because of the uncertainty principle of time and frequency, to control the aspect ratio is not easy. The PSF is changed by the parameters, such as a frequency-range, a frequency-resolution, a time-range, a window function, a sampling frequency, etc. I propose the control method that keeps the aspect ratio of PSF constant with expansion and contraction of an image.