Automated Screening for Speech Disorders Using Acoustic Landmark Detection (2017)

Marisha Speights, Keiko Ishikawa, Joel MacAuslan, Suzanne Boyce
The purpose of this study is to characterize differences in the speech of children with and without speech disorders by using an automated acoustic landmark based approach.
This pilot study explores entropy as a tool for characterizing differences in the landmark (LM) acoustic sequence between normal adults and children with and without a Speech Sound Disorder (SSD). Shannon’s Entropy and ROC analysis are used to evaluate the landmark patterns as potential diagnostic measures of atypical speech production. We discuss these results and our future work toward developing a fully automated clinical screening tool.

Introduction: Speech Disorders in Children

  • Speech requires precision in 1) planning and execution of articulatory targets and 2) sequencing the timing, direction, and force of the articulators. Speech is susceptible to decreased accuracy and precision due to the complexity of such movements 1,2
  • Most children effortlessly learn how to coordinate movements for normal speech production.2-5
  • About one in twelve preschool-aged children, however, show delays in speech production capability that may put them at risk for academic and behavioral difficulties, if not identified and treated.6

Read More… Download PDF

What Are Acoustic Landmarks, and What Do They Describe?

In speech acoustics, landmarks are patterns that mark certain speech-production events. Speechacoustic
landmarks come in two classes: peak and abrupt.

Peak: At present, the peak landmarks detected in SpeechMark® are vowel landmarks (VLMs) and
frication landmarks. These are identified as instants in an utterance at which a maximum (or peak) of
harmonic power or of fractal dimension occurs, respectively, and may be considered the centers of
the vowels or fricated intervals (resp.).

Read More… Download PDF

Peak Landmarks in SpeechMark

Landmarks (LMs) are acoustically identifiable points in an utterance. They come in the form of abrupt
transitions (abrupt LMs) and peaks (peak LMs) of some contour or contours. Here we describe the peak
set of landmarks used in SpeechMark®.

Read More… Download PDF

Spectrogram SNR and Spectrogram Displays

We are often presented with a waveform or spectrogram for which it is helpful to suppress details in noise-dominated sections of time (in the waveform) or of time-frequency(“T-F”, in the spectrogram).

In keeping with the knowledge-based focus of SpeechMark, we are particularly interested in solutions based on broad principles rather than ones that must be determined in a subtle, complicated, or ad hoc fashion, whether by the user or by SpeechMark.

Read More… Download PDF

The SpeechMark MATLAB Toolbox, Cross-Platform Edition

SpeechMark Product Release Notes

SpeechMark Product: The SpeechMark MATLAB Toolbox, Cross-Platform Edition
Operating Systems Supported: All supporting MATLAB v.2017b and later
Product Version: 1.5
Public Release Date : 2024-4-8

———————————————————————————
Installation Notes–PLEASE READ
===============================

1) This product is a standard MATLAB toolbox. To use it, a valid instance of MATLAB (version R2017b or newer) must be installed, as well as a valid version of the MATLAB Signal Processing Toolbox. A valid version of the MATLAB Image Processing toolbox, if present, will be used for certain performance enhancements but is not required.
2) This version includes all code and documentation for v1.5 of 2024-4-8.

———————————————————————————
Resolved Problems:
1) Fixed multiple defects in ‘landmarks’ with over-aggressive voicing and “+g-v”.
2) Improve fallback rule for ‘nonbreathy’.
3) Fixed defects in ‘pltnz’ and ‘plot3vec’ for white-background color changes.
4) Bug fixes in ‘colormap_std’, ‘best_formant_list’, ‘estnoisegram’, ‘estnoisesig’, ‘formant_decay_limits’, ‘mat_conslms’, ‘mat_streamlms’, ‘mat_vowel_segs’, ‘sortperm’, ‘vowel_segs_full’, ‘vowel_segs_std’, and ‘wavelyze_formants’.
5) Help is overhauled in ‘abrupt_events’, ‘abrupt_lms’, ‘landmarks’, ‘lm_abfeatures’, ‘lm_allfeatures’, ‘lm_closant_relndxs’, ‘lm_codes’, ‘lm_duration’.
6) Removed the error function from the Toolbox as it was overriding Matlab’s error function.
7) Replaced calls to ‘wavread’ by ‘audioread’ in some Toolbox functions.
8) Fixed case-sensitive defects in functions starting with DTChk_ .
9) Bug fixes in ‘plot_vowelarea’ and ‘plot_quad_limits’.

———————————————————————————
Enhancements:
1) Several enhancements in ‘vowel_segs_nsil’ equivalent to ‘vowel_segs_full’.
2) Added new function ‘deg_unvstop_asp16’, which gives the correctness of unvoiced, aspirated stop.
3) Optimized the running-time of ‘formant_freq_ratio’, ‘formant_decay_limits’, ‘check_formant12_quad’, ‘check_formant13_quad’, and ‘best_formant_list’.
———————————————————————————
Other Changes:
=====
1) Installer does not require the SpeechMark folder to be named ‘Contents’.
2) Added two new demos, smdemo3 and smdemo4.
———————————————————————————
Known Defects and Limitations:
1) xl_* functions do not support tables.
2) MATLAB system function BUILDDOCSEARCHDB is not used for SpeechMark documentation.

Improving the Accuracy of Automatic Detection of Emotions From Speech

Reza Asadi, Harriet Fell
Computers that can recognize human emotions could react appropriately to a user’s needs and provide more human like interactions.

Some of the applications of emotion recognition:

  • Diagnostic tool for medical purposes
  • Onboard car driving systems to keep the driver alert if stress is detected[1]
  • Similar system in aircraft cockpits
  • Online tutoring

Our contributions:

  • Use new combinations of acoustic feature sets to improve the performance of emotion recognition from speech
  • Provide a comparison of feature sets for detecting different emotions

Read More… Download PDF

Measurement of Child Speech Complexity Using Acoustic Landmark Detection

Keiko Ishikawa, Marepalli B. Rao, Suzanne Boyce
Dysphonia negatively affects speech intelligibility especially in the presence of background noise; however, no clinical tool exists to measure this deficit. Landmark (LM) analysis may serve as the basis of such tool.

The analysis identifies characteristic patterns of abrupt changes in the speech signal over time, and assigns them particular “landmarks.” Consequently, it describes speech as a sequence of LMs.

Read More… Download PDF

Measurement of Child Speech Complexity Using Acoustic Landmark Detection

An important measure of intelligibility in young children is the ability to articulate complex syllables1-4. The development of well-formed syllables in infancy has been shown to be a significant predictor of later communication skills. 1-4 Children with delayed speech acquisition do not show this same developmental trend, and deviations in syllable acquisition may serve as a diagnostic marker of future speech delay.

Read More… Download PDF

Deep Brain Stimulation May Contribute to Dysarthria in Patients with Parkinson’s Disease as Detected by Objective Measures

Craig Van Horne M.D Ph.D, Joel MacAuslan Ph.D, Karen Chenausky M.S CF-SLP, Carla Massari
Dysarthria is found in approximately 80% of patients with Parkinson’s Disease (PD) and significantly limits communication as the severity worsens. Surgical implantation of deep brain stimulators (DBS) into the subthalamic nucleus (STN) has become more common and is an effective treatment for the motoric symptoms of PD. However, the effect of DBS on speech is equivocal.

We have developed computer algorithms that quickly and objectively analyze the speech of PD patients, allowing clinicians to assess the effect of speech on DBS programming or other therapies.

Read More… Download PDF

Spontaneous Vocalization Change in Infants with Severe Impairments using visiBabble

Harriet Fell, Joel MacAuslan, Cynthia J. Cress, Cara Stoll, Kara Medeiros, Jennifer Rosacker, Emily Kurz, Jenna Beckman
Children with difficulty producing speech sounds can practice sounds in play, even prelinguistically. visiBabble is a prototype computer-based program that responds with customized animations to targeted types of infant vocalizations. The program automatically recognizes acoustic-phonetic characteristics of the vocalizations and can selectively respond to utterances with varying levels of complexity (e.g. multisyllable utterances).

This poster reports syllable production changes of three children with physical and speech impairments, ages 1-4, in response to visiBabble reinforcement. Results include immediate effects of visiBabble reinforcement on infant vocalizations as well as longer-term effects of home visiBabble practice on spontaneous sound production.

Read More… Download PDF