Advertisement

Effect of Number of LPCs and Formant Deletion Methods of Inverse Filtering on Acoustic Parameters of Voice

Conference paper
  • 414 Downloads
Part of the Lecture Notes in Mechanical Engineering book series (LNME)

Abstract

Acoustic analysis is one of the efficient, non-invasive, and quantitative methods of voice assessment. The estimation of glottal flow parameters using acoustic analysis is achieved through the method of Inverse Filtering (IF). This study aims at describing effects of the two available methods of inverse filtering—Formant-Based Inverse Filtering (FBIF) and Linear Prediction-Based Inverse Filtering (LPBIF) on acoustic parameters. The effects of formant deletion and number of LPCs on the various vocal parameters—Fundamental frequency based, intensity based, perturbation based, and noise-based measures—were studied. Phonation samples of /a/ at a comfortable pitch and loudness by 30 healthy participants (15 males and 15 females) were recorded on to a PC in a noise-free environment. In the LPBIF-based method, the recorded voice sample was analyzed under five different conditions, i.e., by varying the number of LPCs. The standard value of LPC used in the Vaghmi software is 18. However, in the current study, the number of LPCs was set to 14, 16, 18, 20, and 22, respectively in each of the analysis conditions with other settings remaining the same. In the FBIF-based method, the recorded voice sample was analyzed under four different conditions, i.e., by deleting different formants from the sample. Formant deletion was accomplished using manual methods. The four conditions used were deletion of F1, deletion of F2, deletion of F3, and deletion of F1, F2, and F3 formants. Acoustic parameters of F0, F0 min, F0 max, I0, I0 min, I0 max, JF0, JT0, RAP3, RAP5, SHdB, APQ5, APQ11, HNR, and GNE were compared in all the different conditions mentioned above. A non-parametric Friedman’s test of differences among repeated measures was conducted. The results revealed no change in the value measured across all conditions in both FBIF and LPBIF methods for the parameters of F0, F0 min, F0 max, I0, I0 min, and I0 max. Significant differences across the four formant conditions under FBIF method were found on all perturbation and noise-related measures, i.e., JF0, JT0, RAP3, RAP5, SHdB, APQ5, APQ11, HNR, and GNE. Significant differences across the five LPC conditions were found only on the parameters SHdB, APQ5, APQ11, HNR and GNE. FBIF could be assumed to be more sensitive than LPBIF as the differences across conditions in FBIF were significant in all measured parameters unlike LPBIF where the differences were significant in only five of the parameters. The pros and cons of using each of the two methods for acoustic analysis of voice are discussed in the current study. Further research needs to be done to investigate the effect of varied parameters of inverse filtering in disordered population.

Keywords

Acoustic analysis Inverse filtering Vaghmi Voice LPC Formant-based inverse filtering 

References

  1. 1.
    Amir O, Wolf M, Amir N (2009) A clinical comparison between two acoustic analysis softwares: MDVP and Praat. Biomed Signal Process Control 4(3):202–5CrossRefGoogle Scholar
  2. 2.
    Miller RL (1959) Nature of the vocal cord wave. J Acoust Soc Am 31:667–679CrossRefGoogle Scholar
  3. 3.
    Fant G Acoustic theory of’ speech production, s’ Gravenhage, 2ndGoogle Scholar
  4. 4.
    Atal BS (2006) The history of linear prediction. IEEE Signal Process Mag 23(2):154–161CrossRefGoogle Scholar
  5. 5.
    Atal BS, Schroeder MR (1979) Predictive coding of speech signals and subjective error criteria. IEEE Trans Acoust Speech Signal Process 27:247–254CrossRefGoogle Scholar
  6. 6.
    Itakura F, Saito S (1970) A statistical method for estimation of speech spectral density and formant frequencies. Trans IECE Jpn 53-A:36–43Google Scholar
  7. 7.
    Deliyski DD, Shaw HS, Evans MK (2005) Influence of sampling rate on accuracy and reliability of acoustic voice analysis. Logop Phoniatr Vocology 30(2):55–62CrossRefGoogle Scholar
  8. 8.
    Deliyski DD, Shaw HS, Evans MK, Vesselinov R (2006) Regression tree approach to studying factors influencing acoustic voice analysis. Folia Phoniatrica et Logopaedica 58(4):274–88CrossRefGoogle Scholar
  9. 9.
    Burris C, Vorperian HK, Fourakis M, Kent RD, Bolt DM (2014) Quantitative and descriptive comparison of four acoustic analysis systems: vowel measurements. J Speech Lang Hear Res 57(1):26–45CrossRefGoogle Scholar
  10. 10.
    Ananthapadmanabha TV (2010) Voice analysis using Vaghmi diagnostics software: case studies inverse filtering, voice and speech system, BangaloreGoogle Scholar
  11. 11.
    Vallabha G, Tuller B, Slifka J, Manuel S, Matthies M (2004) Choice of filter order in LPC analysis of vowels. From Sound Sense 50:B148–63Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2021

Authors and Affiliations

  1. 1.Department of Speech Language PathologyAll India Institute of Speech and HearingMysuruIndia
  2. 2.Speech Language Pathologist, Jawahar Lal Nehru Medical College and HospitalAjmerIndia
  3. 3.Department of Speech-Language SciencesAll India Institute of Speech and HearingMysuruIndia

Personalised recommendations