Sliding a window by the hop length and partitioning input signal into window length interval.

 

 

 

 

 

2. Signal Processing

The partitioned input signals typically overlap by the difference between the window and hop sizes.

This overlap ensures temporal continuity and seamless frequency tracking.

Following the application of a window function to each segment, a FFT is proceeded to map the signal into the time-frequency domain.

 

 

Window Function(HANN, Hamming, or etc…,)

 

 

FFT after applying window function on each time-segmented input signal → Time-Frequency Spectrum

 

 

 

 

 

 

3. Representation

The result of STFT is represented as a spectrogram, which shows how the magnitude of frequency components varies over time.

The color intensity represents the energy level at each time and frequency.

Unlike a standard FFT, the spectrogram includes a time axis where each point corresponds to the individual time-segmented intervals.

 

 

Spectrum at each time		Spectrogram
		The intensity is the energy level at each time and  frequency.

 

 

 

 

 

 

4. Interpretation

At t2 and t5, vertical broadband bursts appear across multiple Doppler frequencies, indicating short-duration posture adjustments and slight body movements.

At t3, the energy around the f3 band becomes stronger and wider, corresponding to a deep breathing event with increased respiratory motion.

Meanwhile, the continuous components at f6 and f2 represent weak heartbeat activity and slow body/posture motion that persist over the entire observation period.

 

 

 

 

 

 

 

 

5. Radar Application

Unlike the FFT, which provides only global frequency information over the entire signal,

the STFT reveals how frequency components evolve over time by applying localized time windows.

This makes STFT particularly effective for analyzing non-stationary signals containing time-varying events

such as speech, radar micro-Doppler motion, biomedical activity, and transient mechanical vibrations.

Its key strength is the intuitive visualization of temporal frequency changes through the spectrogram,

enabling simultaneous observation of both timing and spectral behavior.