Excessive laryngeal tension is a significant factor in a range of voice disorders and a crucial target of therapeutic intervention. Vocal hyperfunction (VH), Parkinson’s disease (PD), and other functional and neurological voice disorders involve primary or secondary laryngeal tension and affect over 65% of individuals with voice disorders (Ramig & Verdolini, 1998). Despite this prevalence, clinical assessment is primarily based on unreliable auditory impressions and manual palpation, since standard acoustic measures are not specific to laryngeal tension.
To address this gap, we have proposed an acoustic estimate of laryngeal tension: relative fundamental frequency (RFF) reflects the degree of baseline laryngeal tension by examining short-term changes in fundamental frequency.
Our efforts and the efforts of others have suggested that RFF differs between individuals with and without VH (Stepp et al., 2010; Roy et al., 2016), PD (Stepp, 2013), and spasmodic dysphonia (Eadie & Stepp, 2013), and even normalizes in individuals with VH after successful voice therapy (Stepp et al., 2011; Roy et al., 2016). In typical speakers, RFF reflects changes in vocal effort and correlates with subglottal pressure and listener perceptions of vocal effort (Lien et al., 2015) as well as indirect measurements of laryngeal tension (McKenna et al., 2016). Finally, changes in RFF are seen in high voice users after a period of prolonged vocal abuse, even though there is no perceptual change in their voice (Heller Murray et al., 2016).
Unfortunately, our work has shown that at least six samples are required to reach a stable estimate of RFF (Eadie & Stepp, 2013). Calculating RFF has required manual interaction with waveform displays to determine individual periods of each of the 10 cycles before and after the consonant. Because of the difficulties in determining the exact onset and offset of voicing, a trained technician was required. The necessary time and expertise has been a critical barrier to validation and clinical adoption.
Thus, we have worked to refine RFF stimuli to ensure optimal within-speaker consistency (Lien, et al., 2015; Lien, et al., 2014; Lien & Stepp, 2014) and have developed and validated automated algorithms for its estimation (Lien, 2015; Lien, et al., In Review).
Stimuli: “Uniform utterances” (rather than running speech) with /f/ and /ʃ/ have the lowest within-speaker standard deviations (Lien et al., 2014). Thus, we recommend collection of the following nine RFF “instances” in each participant: /afa afa afa ifi ifi ifi ufu ufu ufu/. Stimuli in this form are compatible with our automated algorithms described below. An audio example of this stimuli is available here:
Version 1 Algorithms: Dr. Yu-An (Stephanie) Lien led the development of these algorithms as part of her dissertation in Biomedical Engineering at Boston University. You can read detailed descriptions of the algorithms in her dissertation. A shorter form of the validation of these algorithms can be found here:
Lien Y.S., Heller Murray E.S., Calabrese C., Michener C.M., Van Stan J., Mehta D.D., Hillman R.E., Noordzij J.P., Stepp C.E. “Validation of an algorithm for semi-automated estimation of relative fundamental frequency”, Annals of Otology, Laryngology, and Rhinology, 126(10), Pg 712 – 716, 2017.
The National Institute on Deafness and Other Communication Disorders (NIDCD) provided the majority of funding for this research. Please feel free to use these version 1 algorithms in scientific research! If you do so, we ask that you cite their use in this way: “RFF values were calculated using an automated MATLAB program; algorithm details can be found in Lien (2015) and validation details can be found in Lien et al., In Press.“ Our lab is continuing to refine these algorithms, so we anticipate new versions to be available in the future. Feel free to leave feedback here.
We do not recommend use of these algorithms or RFF as clinical outcome measures at this time. However, we hope to discover with the use of these faster algorithms whether there is a role for RFF in clinical voice assessment. These algorithms have been applied thus far to study the relationship between RFF and laryngeal tension (McKenna et al., 2016), whether RFF is sensitive to vocal fold surface hydration and vocal loading (Fujiki et al., 2016), and whether RFF can discriminate between differences in vocal function in hyperfunctional voice disorders with and without vocal lesions (Heller Murray et al., 2017).