O-Solution Time-Frequency Analysis Function OS-0527
The user can perform a Fourier transform at any point (frame length and interval). Therefore, since the user can arbitrarily set the extraction time length, this is an effective method for observing very short-term spectral changes.
This analytical method enables the simultaneous analysis of the temporal and spatial variations of complex waveforms, such as sudden or non-stationary acoustic and vibrational phenomena. This method varies the analysis duration depending on the frequency. Because it offers a good balance between time and frequency, it is effective in capturing the overall analysis results.
This is an example of analyzing the impact sound of a golf club, where the frequency changes rapidly in a short period of time.
Because the time duration of the target signal is short, fine spectral changes in the time direction are blurred in FFT analysis (see figure below).
Using STFT, even with the same frame length, the user can arbitrarily set the frame interval and window function length, making it possible to clearly display the time evolution of rapidly changing frequency components while maintaining frequency resolution (see figure on the right).
This is an example of analyzing very short abnormal noises contained in machine noise.
Because the frequency components of the abnormal sound span a wide range, FFT analysis is used to obtain resolution for the lower frequency components.
The frame length needs to be sufficiently large.
However, because the duration of the abnormal noise is very short, it is also necessary to make the frame length as short as possible.
Both frequency resolution and temporal resolution are not sufficiently obtained (left figure).
In this example, the components in the area circled in red (low-frequency components) that would be overlooked in FFT analysis become visible (right figure).
Thus, the wavelet transform analyzes higher frequencies (where the frequency resolution is sufficient) by reducing the frequency resolution and increasing the time resolution, while lower frequencies (where the time change is gradual) by reducing the time resolution and increasing the frequency resolution.
In other words, the analysis method is a stoichiometric filter bank analysis similar to the 1/N octave band analysis commonly used in acoustic analysis.
