Untersuchungen zur Vergleichbarkeit von Ergebnissen der Stimmschallanalysekommerziell verfügbarer Geräte

Communication in our society has become progressively important, so that voice as an important instrument for verbal communication is demanded in-creasingly. Inevitably, the increased strain leads to a greater number of vocal dysfunctions and thus to a growing demand for voice diagnostics. There are various diagnostic methods available which are listed essentially in the basic protocol of the European Laryngological Society (ELS). In addition to the per-ceptual and subjective assessment of voice function, the objective assessment in accordance with computerized voice sound analysis has been very success-ful in recent decades. To achieve this, for example, recorded voice signals are investigated on various parameters. The aim of this study was to examine two commercially available analyzers (lingWAVES by Wevosys, Forchheim and rpSzene® by Rehder / Partner, Ham-burg) on the comparability of its results (vocal acoustic parameters). For this project during the period of February 2009 until June 2010 60 study participants – including 30 voice patients and 30 subjects aged 18 to 81 years – were inves-tigated with vocal sound analysis. The following six vowels [a] [ae], [e], [i], [o] and [u] were phonated as synchronously sustained vowels in the two vocal sound analyzer and were analyzed separately on the basis of the three vocal acoustic parameters jitter, shimmer and Glottal-to-Noise Excitation Ratio. Sub-sequently, the results were evaluated statistically by using Microsoft Excel© – Redmond, U.S., and Bland-Altman plots in the program MedCalc® – Mariakerke, Belgium. The evaluation showed no comparability of lingWAVES and rpSzene® for the investigated vocal parameters jitter, shimmer, or GNE. Jitter showed a large variation in dependence to the "true" value, so that a clear statement with a view to the equivalence of both vocal sound analyzers was just as impossible as a possible assessment regarding the clinical relevance of the differences between both analyzers. Only after modulation of the Bland-Altman plots by calculating the ratios could Shimmer be evaluated. Not only did it show a sys-tematic, but also a clinically relevant difference of five vowels (except for vowel [o]). The determined conversion factors in the form of average ratios however must be evaluated as too vague for the practical application because of the fairly wide limits of agreement. Although the analysis of the GNE-values showed classical Bland-Altman plots with typical distribution of points and a clearly bet-ter agreement than the previous parameters, it was however impossible to give an ultimately clear statement about the comparability of lingWAVES and rpSzene® because of the location of the interval limits of agreement. Because these results correspond with results of other studies and because essential factors, which are partly unresolved, can be found as a part of causali-ty research, computer-based vocal sound analysis cannot be defined as such a reliable and useful tool for the evaluation of voice disorders at the present time, as expected. It is recommended to review the quality of all the individual com-ponents, the standardization of the algorithms and the revision of the popular voice analysis parameters in the computer-assisted analysis of vocal sound.