You're reading an old version of this documentation. If you want up-to-date information, please have a look at 0.10.1.
- time series
Typically an audio signal, denoted by y, and represented as a one-dimensional numpy.ndarray of floating-point values. y[t] corresponds to amplitude of the waveform at sample t.
- sampling rate
The (positive integer) number of samples per second of a time series. This is denoted by an integer variable sr.
A short slice of a time series used for analysis purposes. This usually corresponds to a single column of a spectrogram matrix.
A vector or function used to weight samples within a frame when computing a spectrogram.
- frame length
The (positive integer) number of samples in an analysis window (or frame). This is denoted by an integer variable n_fft.
- hop length
The number of samples between successive frames, e.g., the columns of a spectrogram. This is denoted as a positive integer hop_length.
- window length
The length (width) of the window function (e.g., Hann window). Note that this can be smaller than the frame length used in a short-time Fourier transform. Typically denoted as a positive integer variable win_length.
A matrix S where the rows index frequency bins, and the columns index frames (time). Spectrograms can be either real-valued or complex-valued. By convention, real-valued spectrograms are denoted as numpy.ndarrays S, while complex-valued STFT matrices are denoted as D.
- onset (strength) envelope
An onset envelope onset_env[t] measures the strength of note onsets at frame t. Typically stored as a one-dimensional numpy.ndarray of floating-point values onset_envelope.
Also known as pitch class profile (PCP). Chroma representations measure the amount of relative energy in each pitch class (e.g., the 12 notes in the chromatic scale) at a given frame/time.