Ctrl+K

Audio Classes#

Audio data are at the core or pyfar and three classes can be used for storing, processing, and visualizing such data

  • The Signal class can be used to store equidistant and complete signals that can be converted between the time and frequency domain.

  • The TimeData and FrequencyData classes are intended for incomplete or non-equidistant audio data in the time and frequency domain that can not be converted between the time and frequency domain.

All three classes provide methods for accessing the data and useful meta data such as the sampling rate or number of channels, and almost all functions in pyfar take audio classes as their main input. See audio classes for a complete documentation.

Data inside audio classes: cshape , cdim and caxis

All audio classes can store multi-dimensional data. For example 3 by 4 channel audio data with 128 samples of data per channel would be stored in an array of shape 3 by 4 by 128, also written as (3, 4, 128). In audio signal processing, operations are often performed on channels. To handle this conveniently, pyfar defines the channel shape - or cshape in short - which ignores how many samples or frequency bins are stored in an audio object. In the above example the cshape would be (3, 4) and not (3, 4, 128).

In analogy, the channel axis - or caxis in short - refers to the dimension or axis of the cshape. For example caxis = 0 refers to the first dimension of size 3, and caxis = -1 refers to the last dimension of size 4 and not to the last axis of the data array of size 128. Note that caxis can also be provided as a tuple or list, e.g., caxis = (0, 1) refers to both channel axes of size 3 and 4.

Lastly the channel dimension - or cdim in short - is the length of the cshape.

Signal Types

The Signal class distinguishes between two kinds of signals

  • energy signals are of finite length and energy, such as impulse responses.

  • power signals are finite samples of signals with infinite length and energy, such as noise signals or sound textures.

The difference is important for

  • plotting the frequency response of signals because different signal types require different FFT normalizations.

  • performing arithmetic operations because not all signal types can be combined with each other.