Beginners guide to Praat

Sound editor: Spectral slices (FFT)


[Contents][Spectral analysis][FFT slices]

Select a topic or read through:

Preliminaries

  • Click here to go directly to a quick guide for spectral analysis.
  • Spectral slices (or cross-sections) show the amplitude/frequency spectrum at a selected moment in the signal. They are useful as aids to comparing local spectral events or measuring spectral properties such as formant frequencies, levels or bandwidths. You should already be familiar with spectrograms and formants.

How Praat makes slices

  • Praat offers two analysis methods: FFT and LPC. This section describes FFT slices. Click here for LPC slices.
  • Praat offers two procedures for FFT slices: creation in the Sound editor and creation in the Objects window. This beginner's guide concentrates on the Sound editor procedure and then gives a brief outline of the Objects window procedure. The preliminaries are the same for both procedures - having a signal to work on, understanding the parameter settings, deciding where to make the slice etc.

Example signals used to illustrate FFT slices

  • A male and a female speaker are used to illustrate the creation of spectral slices in the Sound editor. The first is a Swedish adult male speaker saying  finns det dokumentära inslag ("there are documentary elements"):

  • The second is a Swedish adult female speaker saying ett forskningsprojekt ("a research project"):

Getting started

  • Load your signal into the Objects window, select it and click Edit. The Sound editor opens, showing the waveform of your signal along with any analyses that you had used on the previous occasion you had the Sound editor open (Praat remembers your analysis selections from one session to the next).
  • It is useful to have the spectrogram on view (you can make slices from the waveform only, but it is extremely helpful to find your way about in the spectrogram). If the spectrogram is not already visible, open the Spectrum menu and tick Show spectrogram:

  • Turn any other analyses off if they happen to be visible. There is a long and a short way to do this. The long way is to open each relevant analysis menu in turn, and untick it (Pitch, Intensity, Formant, Pulses as the case may be). The short way is to open the View menu and select Show analyses.

  • Then, in the dialog box that appears, untick the unwanted analyses (and in fact you could also have selected the spectrogram here too, and fixed all these selections and deselections in one go):

  • Note the setting for Longest analysis. The spectrogram will be displayed only when the length of signal on view in the Sound editor is no longer than this setting. If the portion of the signal on view is longer than the Longest analysis setting, you will get a warning. If you get this warning, zoom in to a shorter portion of the signal, until the spectrogram appears.

Parameter settings

  • The parameter settings for spectral slices are the same as those already set for the spectrogram. If you change any settings now for the slice, the spectrogram will immediately be modified too. To check the settings, open the Spectrum menu and select Spectrogram settings. The following dialog appears, showing your current settings. Make any adjustments you need (the parameters are explained as they arise below). To restore the default settings, click Revert to standards.

Wideband and narrowband slices.

  • Wideband slices show the formant structure at the selected location, narrowband slices show more detailed spectral structure (which will be the voice harmonics if you are examining a voiced segment). This works the same way as for spectrograms. The filter bandwidth is set at Window length (s) in the Spectrogram settings dialog box. The unit for this setting is seconds and not Hz, and is the time constant of the filter rather than the bandwidth.
  • For wideband slices, a time constant of 4ms or 5ms will generally work well (but remember it has to be entered in the settings box in seconds, i.e. 0.004 or 0.005). The default setting is 0.005s, which is fine for most adult male voices, but might be on the large side at slice locations where the voice fundamental is relatively high (especially for men with high pitched voices, for most women, or for children). In this situation the voice harmonics might intrude. If this occurs, a slightly smaller time constant (0.004 or 0.003s) might be more suitable. Experiment to find the best setting for each voice and each slice location.
  • For narrowband slices, set Window length to 0.03s, which will be fine for both male and female voices.
  • The filter time constant also defines the duration of the slice, such that a 5ms time constant gives a 5ms wide slice, and so on. This also means that a 5ms wideband slice may well comprise less than one glottal pulse, while a 30ms narrowband slice may comprise three or more glottal pulses. Remember that Praat balances the slice evenly around the cursor location, so that you know precisely what part of the signal is included. Here are examples of wideband and narrowband slices from [i]-like vowels by the male and female speaker respectively. These can be compared with similar examples created in the Objects window at the same location.

 

 

  • The same utterance contains an instance of the difficult situation presented by a high fundamental frequency. In the final vowel, the fundamental rose to 326Hz and a widband slice taken there with the default 0.005s window showed intruding voice harmonics that confuse the identification of the formants:

  • Clearly, the fundamental frequency is so high in this example that the formants appear to coincide with voice harmonics. A slightly smaller setting of Window length will sometimes help exclude such intruding harmonics. Reducing the window from the default 0.005s to 0.003s gives the following:

  • This is an improvement, but there are still ambiguities. The first and second harmonics still intrude and make it difficult to identify F1 properly, and the eleventh and twelfth harmonics still intrude and make F4 difficult to identify properly. In this situation, an LPC slice is often less ambiguous:

Windowing funtion

  • This is selected from the dropdown menu at Window shape in the Spectrogram settings dialog. The default, and recommended, window function is Gaussian.

Dynamic range

  • This determines how far background noise is allowed to intrude and show up in the spectrum diagram. Noise is not usually a problem if your signal has been recorded carefully in silent conditions. Typical background noise comes from room ventilation and air conditioning, road traffic outside, aircraft passing overhead, computer fans, other voices nearby. The default dynamic range is 50dB (the weakest sound pressure displayed in the spectrum diagram is 50dB below the strongest sound pressure). Reduce this number in 3db steps to exclude noise energy (as a rule, once you have set this for the spectrogram you can keep the same setting for the slices).

Where to make your slice

Analysis frame location

  • You select a location for the slice by clicking in the Sound editor window. But Praat actually takes the spectrum at the nearest spectrogram analysis frame. This frame location is determined by Number of steps in the Spectrogram settings dialog. The default setting is 1000 steps, which gives a step size of 2.5ms if you have 2.5s of signal on view in the Sound editor window, and in such a situation the slice will be taken at anything up to 1.25ms away from your cursor location. However, it is good practice to zoom in around the slice location in order to position the cursor more precisely. The 1000 steps are then spread over the zoomed area, for example if you are viewing 250ms of signal after zooming, the step size will be 0.25ms and the difference between cursor location and analysis frame location will have become negligible.

Deciding where to make the slice

  • For now, while you are practicing the procedures and parameter settings, it might not matter where you make a slice. But for serious work, the location of the slice will be defined uniquely depending on your experiment design. Remember you should zoom in to magnify the area you are interested in, and you can specify an exact location by opening the Select menu and selecting Move cursor to. Here is an example, to the left a brief portion of a signal showing a sequence of a vowel and two consonants [..ins..] with a small selection during [i], and to the right that same selection zoomed to one glottal pulse, allowing the cursor to be positioned more precisely:

  • Remember also that the waveform, and hence the spectrum, of a speech signal is continuously changing. Your slice will be misleading if you miss the intended location by a few milliseconds, or if your intended location is not properly defined.
  • The energy distribution also changes within a glottal pulse. With low pitched male voices (glottal pulses of 10ms or more), it will be possible to make a number of wideband slices within one glottal pulse and each will be different. The energy you see in voiced parts of a spectrogram (such as to the left in the previous example) comes from the stronger first part of each glottal pulse, hence the typical vertical lines on the spectrograms. The lighter gaps between these vertical lines portray the weaker endings of each glottal pulse. This is illustrated in the next example. The next diagram shows the zoomed waveform of part of an [i]-like vowel, with one period shaded, representing one glottal pulse. The two vertical lines A and B mark the locations of two wideband slices 5ms apart so that they will not overlap. A is in the stronger early part of the pulse, B is in the weaker latter part of the pulse.

  • The next diagram compares the two broadband slices (5ms Gaussian) taken at the positions A and B respectively, illustrating this difference:

  • This demonstrates (i) that the spectrum varies within a glottal pulse, (ii) a wideband slice can pick up that variation, (iii) an imprecision of a few milliseconds can pick up that variation. This emphasizes the need for careful definition of what you are looking for when making a slice, careful definition of a slice location in order to find what you are looking for, and (iii) careful selection of parameter settings in order to see what you are looking for. To see the spectrum of this glottal pulse, it would be advisable to set the filter duration equal to the length of the glottal period, i.e. around 10ms.

Creating a slice in the Sound editor by FFT

  • This method creates a slice directly from the Sound editor and displays it in a new window.
  • Load your signal, open the Sound editor, have the spectrogram on view to help you find your way in the signal, adjust any analysis settings as necessary, all as outlined above.
  • Position the cursor precisely where the slice is to be taken, zooming as necessary, and taking advantage of Move cursor to in the Select menu.
  • In the Spectrum menu, click View spectral slice:

  • The slice appears in a new window.
  • At the sime time a Spectrum object named slice is placed in the Objects list in the Objects window. You can rename it (a good idea if you are making more slices as part of your work).

Creating an FFT slice in the Objects window

  • This method requires the spectrum location to be extracted as a Sound object to the Objects list, and then an FFT is performed on it there. The resulting spectrum can be viewed from the Objects window.
  • In the Sound editor, position the cursor in the waveform or spectrogram where you want the slice (see above).
  • Make a selection at that location, evenly balanced round the cursor position. This is most easily done from the Select menu, clicking Select and entering the endpoints of the selection. The duration of the selection defines the filter, e.g. 5ms for a wideband slice, 30ms for a narrowband slice. In this example the cursor is located at 0.133s and a 5ms selection is being balanced around it by entering the endpoints 0.1305 and 0.1355:

  • Now extract the selection. Open the File menu, click on Extract windowed selection:

  • and choose a windowing method from the list that is offered:

  • The default window function is Hanning. The extracted selection has the default name slice. This can be renamed here, or later in the Objects window. The extracted selection is listed as a Sound object in the Objects window, with the name given above, i.e. your new name or the default name slice.
  • Select this object, and then open the Spectrum menu in the list of actions to the right, and choose To Spectrum (fft):

  • The FFT is calculated and a Spectrum object is placed in the list, with the same name as its Sound object. This object can be renamed.
  • Select the Spectrum object and click the Edit button:

Printing and saving FFT slice diagrams

  • FFT slices are printed from the Picture window, where they are transferred from the Objects window. How to do this is outlined in detail elsewhere, but a necessary first step is to get the slice into a Spectrum object in the Objects list. This was outlined just above, both methods putting a Spectrum object in the Objects list.
  • Then mark out the area the FFT slice is to occupy in the Picture window. Position the mouse pointer where you want the top left corner, then hold the left mouse button down and drag the pointer to where you want the bottom right corner. Then release the mouse button.
  • Then go back to the Objects window, select the Spectrum object, and click the Draw button:

  • The Draw spectrum dialog appears where you can adjust settings that affect the appearance of the spectrum diagram, then click O.K. and the slice appears in the selected area in the Picture window:

  • This example was produced using the default Draw spectrum settings.
  • The commands for printing, or writing image files, are in the File menu. Click here for more details on printing and saving FFT slices.

 


[Contents][Spectral analysis][FFT slices]

Praat for beginners - Sidney Wood - 12 May 2004

 Copyright (c) 1995-2004 Sidney Wood. All rights reserved.