Sound

Sound

One of the types of objects in Praat. For tutorial information, see all of the Intro.

Commands

Creation:

• Record mono Sound... (from microphone or line input, with the SoundRecorder)

• Record stereo Sound...

• Create Sound from formula...

• Create Sound as tone complex...

• Create Sound as gammatone...

• Create Sound as Shepard tone...

Opening and saving:

• Sound files

You can also use the text and binary (real-valued) formats for Sounds, like for any other class:

• Save as text file...

• Save as binary file...

Viewing and editing:

• SoundEditor, ManipulationEditor, TextGridEditor, PointEditor, PitchTierEditor, SpectrumEditor

Playing:

• Sound: Play

• PointProcess: Hum

• PointProcess: Play

Drawing:

• Sound: Draw...

• Sound: Draw where...

• Sound: Paint where...

• Sounds: Paint enclosed...

Queries:

structure:

• time domain

• Get number of samples

• Get sampling period

• Get sampling frequency

• Get time from sample number...

• Get sample number from time...

content:

• Sound: Get value at time...

• Sound: Get value at sample number...

shape:

• Sound: Get minimum...

• Sound: Get time of minimum...

• Sound: Get maximum...

• Sound: Get time of maximum...

• Sound: Get absolute extremum...

• Sound: Get nearest zero crossing...

statistics:

• Sound: Get mean...

• Sound: Get root-mean-square...

• Sound: Get standard deviation...

energy:

• Sound: Get energy...

• Sound: Get power...

in air:

• Sound: Get energy in air

• Sound: Get power in air

• Sound: Get intensity (dB)

Modification:

• Matrix: Formula...

• Sound: Set value at sample number...

• Sound: Filter with one formant (in-place)...

• Sound: Pre-emphasize (in-place)...

• Sound: De-emphasize (in-place)...

Annotation (see Intro 7. Annotation):

• Sound: To TextGrid...

Periodicity analysis:

• Sound: To Pitch (filtered autocorrelation)...

• Sound: To Pitch (raw cross-correlation)...

• Sound: To Pitch (raw autocorrelation)...

• Sound: To Pitch (filtered cross-correlation)...

• Sound: To Pitch (shs)...

• Sound: To Harmonicity (ac)...

• Sound: To Harmonicity (cc)...

• Sound: To PointProcess (periodic, cc)...

• Sound: To PointProcess (periodic, peaks)...

• Sound & Pitch: To PointProcess (cc)

• Sound & Pitch: To PointProcess (peaks)...

• Sound: To Intensity...

Spectral analysis:

• Sound: To Spectrum...

• Sound: To Spectrogram...

• Sound: To Formant (burg)...

• Sound: To Formant (sl)...

• Sound: LPC analysis

• Sound: To LPC (autocorrelation)...

• Sound: To LPC (covariance)...

• Sound: To LPC (burg)...

• Sound: To LPC (marple)...

Filtering (see Filtering tutorial):

• Sound: Filter (pass Hann band)...

• Sound: Filter (stop Hann band)...

• Sound: Filter (formula)...

• Sound: Filter (one formant)...

• Sound: Filter (pre-emphasis)...

• Sound: Filter (de-emphasis)...

• LPC & Sound: Filter...

• LPC & Sound: Filter (inverse)

• Sound & Formant: Filter

• Sound & FormantGrid: Filter

Conversion:

• Sound: Resample...

Enhancement:

• Sound: Lengthen (overlap-add)...: lengthen by a constant factor

• Sound: Deepen band modulation...: strenghten intensity modulations in each critical band

Combination:

• Sounds: Convolve...

• Sounds: Cross-correlate...

• Sound: Autocorrelate...

• Sounds: Concatenate

Synthesis

• Source-filter synthesis tutorial

• Manipulation (overlap-add etc.)

• Spectrum: To Sound

• Pitch: To Sound...

• PointProcess: To Sound (pulse train)...

• PointProcess: To Sound (hum)...

• Pitch & PointProcess: To Sound...

• Articulatory synthesis tutorial

• Artword & Speaker: To Sound...

Inside a Sound

With Inspect, you will see that a Sound contains the following data:

x_min: start time, in seconds.
x_max > x_min: end time, in seconds.
n_x: the number of samples ( 1).
dx: sample period, in seconds. The inverse of the sampling frequency (in Hz).
x₁: the time associated with the first sample (in seconds). This will normally be in the range [xmin, xmax]. The time associated with the last sample (i.e., x₁ + (nx − 1) dx)) will also normally be in that range. Mostly, the sound starts at t = 0 seconds and x₁ = dx / 2. Also, usually, x_max = n_x dx.
z [1] [1..n_x]: the amplitude of the sound (stored as single-precision floating-point numbers). For the most common applications (playing and file I-O), Praat assumes that the amplitude is greater than -1 and less than +1. For some applications (modelling of the inner ear; articulatory synthesis), Praat assumes that the amplitude is expressed in Pascal units. If these interpretations are combined, we see that the maximum peak amplitude of a calibrated sound is 1 Pascal; for a sine wave, this means 91 dB SPL.

Limitations

Since the Sound object completely resides in memory, its size is limited to the amount of RAM in your computer. For sounds longer than a few minutes, you could use the LongSound object instead, which you can view in the LongSoundEditor.

Links to this page

© Paul Boersma 2016-10-13