Why MDS for voice analysis?
                                                 Voice Parameters
                                      Role of MDS for voice analysis
                                                  Published studies

Parameters of voice:

To fully appreciate the use of MDS for voice analysis, one must first recognize the complexity of the human voice:

                         Parameters of Voice

Rate of vf vibration frequency--fo,cps, Hz pitch
Amplitude of vf vibration SPL, intensity--dB loudness
Periodicity of vf vibration (stability) perturbation-jitter,shimmer quality
Complexity of vf vibration  range flexibility

                                                    Graph obtained from class notes:  CD 583  (Leeper, 2001)

The parameters of voice can be considered multidimensional in nature.  Furthermore, "little is known about the way various parameters interrelate to result in the unique and distinctive vocal quality of an individual."                                                                                   (Kemster, 1991)

As a result of the complex nature of the human voice, dramatic differences exist regarding: disciplinary perspectives, levels of training, experience, and available instruments for evaluating voice disorders.

        "There is a wide assortment of variables that are possible targets for quantification.
        Potentially relevant phenomena and indices include, to name a few, Fo, vocal intensity,
        mean glottal airflow, instantaneous glottal volume velocity (via inverse filtering), air
        pressure, vocal fold motions (via stroboscopy), frequency and intensity perturbation,
        vocal fold contact area (via electroglottography), and the laryngeal electromyograms...
       The lack of internal agreement about which of these (or other) measures are most useful,
        either alone or in combination, is particularly glaring and strongly inhibits the application
        of any."                                                                                      (Bless & Baken, 1992)

According to Andrews (1999), there are no existing guidelines to standardize the selection and implementation of the wide variety of vocal function tests.   This warrants the need for further research to ensure the use of reliable methods for determining voice quality and voice disorders.  

        The International Association for Logopedics and Phoniatrics (IALP) established a
        voice committee to examine issues and make recommendations concerning the standard-
        ization of voice evaluation procedures.  This committee has published a survey along
        with report in hopes to stimulate professional dialogue in this area:
        "In the next decade there will undoubetedly be wide discussion and hopefully some
        resolution of critical issues such as calibration requirements, testing environments,
        and examiner qualifications."                                                                (Andrews, 1999)

The Role of MDS

Carroll & Arabie (1980), stated that, "explosive growth" has occurred in the number and variety of models and methods, along with the proliferation of applications of MDS within many different fields.  Unfortunately, "MDS remains an area characterized by a considerable lag between new metholological developments and routine used by nonspecialists."

Considering the "multidimensional" nature of the human voice, consideration of the following points is strongly urged:

        1)  Past research literature for voice parameters is limited and has failed to fully develop
            parameters for organizing concepts and underrlying dimensions.
        2) MDS may provide further detection of patterns and structures that lie hidden in
            large amounts of data.
        3) MDS can promote accurate interpretation and application of data by representing the
            data in forms more accessible to the human eye.  Thus, contributing to our
            further understaning of the human voice.

                       "I think it is an important and underused statistical tool."
                                          Special Interest Division 3 (SID 3) member commenting on MDS
                                                                Deirdre D. Michael, Ph.D., CCC-SLP

Published studies

MDS has (and continues to) contribute to our knowledge of the human voice.

     The following is a list of published studies that employed MDS for voice analysis:

Zraick, Richard I.,  (2000).  Multidimensional Scaling of Nasal Voice Quality.  
     Journal of Speech , Language and Hearing Research, 43:  989-996.  
  • MDS was used  to determine of the perception of nasality is unidimensional or multidimensional, and to identify possible co-occurring dimensions that influence listeners' perception to this quality. 
Kent, Ray, D.  (1996).  Hearing and Believing:  Some limits to the Auditory-Perceptual 
     Assessment of Speech and Voice Disorders.  AJSLP, (5).   
  • Supports the use of MDS as a method for identifying the number and nature of dimensions used in perceptual judgements of voices.
Kreiman, Jody., Gerratt, Bruce. R.  (1996).  The perceptual structure of pathologic voice 
     quality.  J.Acoust. Soc. Am., 100(3).
  • MDS was used to examine the perceptual structure of large samples of pathologic male and female voices, in an attempt to establish a valid set of perceptual features for pathologic voice quality that will generalize to any clinical setting or population.  
Kreiman, Jody.  (1993).  Perceptual Evaluation of Voice Quality:  Review, 
     Tutorial, and a Framework for Future Research.  Journal of Speech and Hearing 
     Research, 36:  (21-40).  
  • Spplied MDS to systematically examine previous research literature concerned with the assessment of speech and voice quality to determine of listeners make reliable judgements about voice quality.  
Pausewang-Gelfer, Marylou.  (1993).  A Multidimensional Scaling Study of Voice 
     Quality in Females.  Phonetica, 50:  15-27.
  • Invistigated the perceptual correlates of dimensions resulting from an MDS analysis of listeners (dis)similarity judgements of normal female voices presented in sentences.  
Kreiman, Jody.  et. al.  (1992).  Individual Differences in Voice Quality Perception.  
     Journal of Speech and Hearing Research, 35: 512-520
  • Monmetric MDS was used to determine: 1)  what (if any) aspects of voice quality were perceptually important across listeners and voice sets, and what characteristics were unique to individual listeners and voice sets; and 2) if differences in listeners' perceptual judgements were systematically related to selected acoustic characteristics of the voices.  
Kemster, Gail. B., Kistler, Doris, J. Hillenbrand, James.  (1991).  Multidimensional 
     Scaling Analysis of Dysphonia in Two Speaker Groups.  JSHR, 34: 534-543.
  • Used MDS to determine the perceptual dimensions related to the quality of dysphonic vowels produced by 30 female talkers.
Kreiman, Jody., Gerratt, Bruce. R., Precoda, Kristen.  (1990).  Listener Experinece and 
     Perception of Voice Quality.  Journal of Speech and Hearing Research, 33: 103-115
  • MDS was used to determine the characteristics of dysphonic and normal voices that are perceptually important for listeners with and without clinical training. 


         Back         Home       Onward



  "What we need are more people who specialize in the impossible."
Theodore Roethke

                                                                                     Cummins - 2001