Quantitative and Descriptive Comparison of Four Acoustic Analysis Systems: Vowel Measurements PurposeThis study examines accuracy and comparability of 4 trademarked acoustic analysis software packages (AASPs): Praat, WaveSurfer, TF32, and CSL by using synthesized and natural vowels. Features of AASPs are also described.MethodSynthesized and natural vowels were analyzed using each of the AASP's default settings to secure 9 acoustic measures: fundamental frequency ... Research Article
Research Article  |   February 01, 2014
Quantitative and Descriptive Comparison of Four Acoustic Analysis Systems: Vowel Measurements
 
Author Affiliations & Notes
  • Carlyn Burris
    University of Wisconsin—Madison
  • Houri K. Vorperian
    University of Wisconsin—Madison
  • Marios Fourakis
    University of Wisconsin—Madison
  • Ray D. Kent
    University of Wisconsin—Madison
  • Daniel M. Bolt
    University of Wisconsin—Madison
  • Disclosure:The authors declare no financial conflict of interest with any of the software systems considered in this review. The authors did contact all software developers to seek clarification or obtain additional information. In particular, they requested and obtained from the developer of TF32 software updates that provided formant bandwidth and fourth formant measurements that enabled the authors to meet the purpose of this study in terms of having comparable measurements to assess across all four software packages.
    Disclosure:The authors declare no financial conflict of interest with any of the software systems considered in this review. The authors did contact all software developers to seek clarification or obtain additional information. In particular, they requested and obtained from the developer of TF32 software updates that provided formant bandwidth and fourth formant measurements that enabled the authors to meet the purpose of this study in terms of having comparable measurements to assess across all four software packages.×
  • Correspondence to Houri K. Vorperian: vorperian@waisman.wisc.edu
  • Editor: Jody Kreiman
    Editor: Jody Kreiman×
  • Associate Editor: Scott Thomson
    Associate Editor: Scott Thomson×
Article Information
Speech, Voice & Prosodic Disorders / Voice Disorders / Speech, Voice & Prosody / Speech
Research Article   |   February 01, 2014
Quantitative and Descriptive Comparison of Four Acoustic Analysis Systems: Vowel Measurements
Journal of Speech, Language, and Hearing Research, February 2014, Vol. 57, 26-45. doi:10.1044/1092-4388(2013/12-0103)
History: Received April 2, 2012 , Revised January 30, 2013 , Accepted May 23, 2013
 
Journal of Speech, Language, and Hearing Research, February 2014, Vol. 57, 26-45. doi:10.1044/1092-4388(2013/12-0103)
History: Received April 2, 2012; Revised January 30, 2013; Accepted May 23, 2013
Web of Science® Times Cited: 7

PurposeThis study examines accuracy and comparability of 4 trademarked acoustic analysis software packages (AASPs): Praat, WaveSurfer, TF32, and CSL by using synthesized and natural vowels. Features of AASPs are also described.

MethodSynthesized and natural vowels were analyzed using each of the AASP's default settings to secure 9 acoustic measures: fundamental frequency (F0), formant frequencies (F1–F4), and formant bandwidths (B1–B4). The discrepancy between the software measured values and the input values (synthesized, previously reported, and manual measurements) was used to assess comparability and accuracy. Basic AASP features are described.

ResultsResults indicate that Praat, WaveSurfer, and TF32 generate accurate and comparable F0 and F1–F4 data for synthesized vowels and adult male natural vowels. Results varied by vowel for women and children, with some serious errors. Bandwidth measurements by AASPs were highly inaccurate as compared with manual measurements and published data on formant bandwidths.

ConclusionsValues of F0 and F1–F4 are generally consistent and fairly accurate for adult vowels and for some child vowels using the default settings in Praat, WaveSurfer, and TF32. Manipulation of default settings yields improved output values in TF32 and CSL. Caution is recommended especially before accepting F1–F4 results for children and B1–B4 results for all speakers.

Acknowledgments
This work was supported by National Institute on Deafness and Other Communication Disorders Research Grant R01-DC 006282 and by National Institute of Child Health and Human Development Core Grant P-30 HD03352. We declare no financial conflict of interest with any of the software systems considered in this review. We did contact all software developers to seek clarification or obtain additional information. In particular, we requested and obtained from the developer of TF32 software updates that provided formant bandwidth and fourth formant measurements that enabled us to meet the purpose of this study in terms of having comparable measurements to assess across all four software packages. We thank Jan R. Edwards and Gary G. Weismer for providing comments on previous versions of this article. We also thank Erin Nelson for assistance with graphing, Ekaterini Derdemezis for verifying revision accuracy, Allison G. Petska for securing references, and Michael P. Kelly for assisting with statistical analysis. This research was originally submitted by Carlyn Burris as a master's thesis for the Department of Communicative Disorders at the University of Wisconsin—Madison. Portions of this research were presented in at the 2011 Annual Convention of the American Speech-Language-Hearing Association Convention in San Diego, CA.
Order a Subscription
Pay Per View
Entire Journal of Speech, Language, and Hearing Research content & archive
24-hour access
This Article
24-hour access