The SpeechMark family of products detects and measures acoustic landmarks —acoustic events that correlate with changes in speech articulation. Landmarks are useful for investigating non-lexical attributes of speech, such as differences in production of the same speech material by the same speaker under different conditions.

The SpeechMark WaveSurfer Plug-in for Windows adds landmark support to the WaveSurfer speech analysis package.

Features

The SpeechMark Plug-in:

• Analyzes speech files to detect and identify the location of acoustic landmarks
• Displays the location of acoustic landmarks on a pane within WaveSurfer
  • Landmark labels are superimposed on the audio signal waveform.

• Displays power contours in selected frequency bands that are used to identify acoustic landmarks

• Supports user preferences for landmark configuration
  • As shown in Figure 3, the set of landmarks to be displayed and output can be modified by the user.

• Generates output files which specify the location (time) and type of the detected landmarks
  • All versions of SpeechMark generate a landmark file (“.lm” file—see Figure 4). A landmark files is a text file that contains one line of information for each generated landmark. That line contains a timestamp, the label of the landmark, and a numerical “strength” value;

• The SpeechMark Plug-in also generates a WaveSurfer “lab.lab” file which represents acoustic landmarks as standard WaveSurfer labels. The user can open this landmark label file in WaveSurfer to use these landmarks as standard WaveSurfer transcription labels.

• Computes a Vowel Space Plot for detected vowel landmarks
  • In the Vowel Space Plot, SpeechMark plots the (F1, F2) positions of detected vowel landmarks.
  • Vowels with unusually high formant bandwidth are plotted in green, the rest are plotted in red.
  • In addition, as a visual guide, SpeechMark also draws a polygon (dashed blue line) showing the approximate range of the adult vowel space. That is, the polygon shows accepted values of the adult vowel space, (not boundaries derived from the vowels being plotted).

System Requirements

• Windows XP or Windows 7
• The latest release of WaveSurfer that can be downloaded from the
  sourceforge.net web site (currently version 1.8.8p4) or WaveSurfer 1.8.5
• The MATLAB Compiler Runtime object (which is normally downloaded
  and installed automatically as part of the SpeechMark product installation process).

Downloads

• Click here to download this SpeechMark product.

The SpeechMark family of products detects and measures acoustic landmarks —acoustic events that correlate with changes in speech articulation. Landmarks are useful for investigating non-lexical attributes of speech, such as differences in production of the same speech material by the same speaker under different conditions.

The SpeechMark MATLAB Toolbox Product adds acoustic landmark support to the MATLAB omputational environment. It comes complete with comprehensively documented code, sample data files, and demo scripts.

Features

The SpeechMark Toolbox:

• Computation of pitch tracks, “cepstrograms,” and band energy envelopes within speech signals;
• Automatic detection of both abrupt and maximal acoustic landmarks in speech;
• Post-processing of generated landmark sequences into a “syllable” structure which reveals something about the complexity of the utterances in the source, including voiced regions (within which pitch measurements are meaningful), and certain “pseudo-speech” regions which can be classified as non-speech noise.
• Generation of landmark-related plots and images;
• Generation of a standard MATLAB .mat (binary) file containing information about detected landmarks, amplitude envelope, and other signal contours and attributes;
• Generation of a standard SpeechMark “.lm” file containing information about detected landmarks;
• Tools for processing real time audio streams and generating acoustic landmarks in real time;
• Comprehensive toolbox documentation compatible with the MATLAB Supplemental Software browser, including a roadmap page, example scripts and output files, and function pages;
• Invocation of the MATLAB Imaging Toolbox, when available, to speed up computations.

System Requirements:

One of the following operating systems:

• Windows 10, Apple OSX (Mountain Lion or later);
• An installed copy of MATLAB (version R2017b or newer);
• An installed copy of the MATLAB Signal Processing Toolbox and, optionally, Image Processing Toolbox.

Potential Enhancements:

• A GUI-based demo, so that the end user can quickly get a feel for the various parameters and how they affect the processing/result;
• A user control that will allow a user to specify features of the landmarks-in-waveform plot to be included or removed.

Screen Shots to be displayed in Product documentation:

• Landmarks-in-waveform plot;
• Vowel space plot.

Downloads

• Click here to download this SpeechMark product.