Frequency resolution

from Wikipedia, the free encyclopedia

The frequency resolution describes the smallest frequency difference between two tones (or sinusoidal processes) that can still be differentiated. It describes the application of the physical term resolution to the dimension of frequency . The term is mainly used in acoustics (especially psychoacoustics ) and signal processing .

On the one hand, the resolution is specified by the quantization of the frequency. The frequency spectrum can be determined with the help of a filter bank (e.g. third octave spectrum). Then the resulting frequency resolution in the entire spectral range is a third octave.

On the other hand, the analysis time determines the resolution. The frequency resolution cannot be significantly better than the reciprocal of the analysis time. This is a property of the discrete Fourier transform , with which frequency spectra are mostly determined. In contrast, the resolution can be improved if additional knowledge is available or assumptions can be made. Then high-resolution methods (e.g. maximum likelihood method or maximum entropy method ) can increase the resolution. Window functions for reducing the sidelobes in the spectrum reduce the resolution.

The ability of humans or animals to distinguish between two tones with slightly different frequencies is different both individually and across the audible range : the relationship between the physical frequency of a tone and its physiologically perceived tone is not linear. This law is shown in the Bark scale .

The extent to which this ability to distinguish between frequencies is decisive for the degree of acoustic orientation ability, for example with regard to directional hearing or the cocktail party effect , which describes speech recognition in the case of loud background and background noises.

Web links