Melspektrogrammipohjaiset

Melspektrogram is a representation of the spectrum of a signal, such as audio, as it changes over time, with the frequency axis scaled according to the Mel scale. The Mel scale is a perceptual scale of pitches judged by listeners to be equal in distance from one another. This scale is non-linear, meaning that equal distances on the Mel scale correspond to increasingly larger distances in actual frequency as frequency increases. This non-linearity is designed to better approximate human auditory perception, where our sensitivity to frequency changes is less precise at higher frequencies.

The process of creating a melspektrogram typically involves several steps. First, the audio signal is divided into short, overlapping frames. Then, a Fourier transform is applied to each frame to obtain its frequency spectrum. This spectrum is then mapped onto the Mel scale by applying a set of triangular filters, often referred to as Mel filters. These filters are spaced logarithmically in frequency, mimicking the non-linear nature of human hearing. The output of these filters, when summed for each frame, provides the Mel-frequency representation. Finally, these Mel-frequency representations are arranged chronologically to form the melspektrogram, which is often visualized as an image where time is on one axis, Mel frequency on the other, and the intensity of the color represents the amplitude of the signal at that time and Mel frequency. Melspektrograms are widely used in speech recognition, music information retrieval, and other audio analysis tasks due to their ability to capture perceptually relevant features of sound.