This study evaluates the practical value of traditional and non-standard acoustic measures for detecting dysphonia in Parkinson's disease (PD) patients. The researchers collected sustained phonations from 31 subjects, 23 of whom had PD, and analyzed 10 highly uncorrelated measures. An exhaustive search of all possible combinations of these measures using a kernel support vector machine (SVM) found that the best combination achieved an overall correct classification performance of 91.4%, distinguishing healthy individuals from PD patients. The study concludes that non-standard methods, when combined with traditional harmonics-to-noise ratios, are most effective in separating healthy from PD subjects. These non-standard methods are robust to variations in acoustic environments and individual characteristics, making them suitable for telemonitoring applications. The study introduces a new measure of dysphonia called Pitch Period Entropy (PPE), which is robust to many uncontrollable confounding effects. PPE is calculated by first obtaining the pitch sequence of the phonation and converting it to the logarithmic semitone scale. A linear whitening filter is then applied to remove linear temporal correlations, producing a relative semitone variation sequence. A discrete probability distribution of these variations is constructed, and the entropy of this distribution is calculated, which characterizes the extent of non-Gaussian fluctuations in the sequence of relative semitone pitch period variations. The study also discusses the importance of feature selection in reducing the number of measures to a manageable size for classification. Correlation filtering was applied to remove highly correlated features, leaving 10 measures. The best classification performance was achieved by combining HNR, RPDE, DFA, and PPE. The results show that PPE alone provides a significant improvement in classification performance, and it appears in all the best performing subsets. The study highlights the limitations of traditional measures in noisy environments and the advantages of non-standard measures in robustness and performance. The findings suggest that non-standard measures are more effective for telemonitoring applications due to their robustness to variations in acoustic environments and individual characteristics. The study also notes that future research should test these findings using voice signals recorded in more typical telemonitoring environments.This study evaluates the practical value of traditional and non-standard acoustic measures for detecting dysphonia in Parkinson's disease (PD) patients. The researchers collected sustained phonations from 31 subjects, 23 of whom had PD, and analyzed 10 highly uncorrelated measures. An exhaustive search of all possible combinations of these measures using a kernel support vector machine (SVM) found that the best combination achieved an overall correct classification performance of 91.4%, distinguishing healthy individuals from PD patients. The study concludes that non-standard methods, when combined with traditional harmonics-to-noise ratios, are most effective in separating healthy from PD subjects. These non-standard methods are robust to variations in acoustic environments and individual characteristics, making them suitable for telemonitoring applications. The study introduces a new measure of dysphonia called Pitch Period Entropy (PPE), which is robust to many uncontrollable confounding effects. PPE is calculated by first obtaining the pitch sequence of the phonation and converting it to the logarithmic semitone scale. A linear whitening filter is then applied to remove linear temporal correlations, producing a relative semitone variation sequence. A discrete probability distribution of these variations is constructed, and the entropy of this distribution is calculated, which characterizes the extent of non-Gaussian fluctuations in the sequence of relative semitone pitch period variations. The study also discusses the importance of feature selection in reducing the number of measures to a manageable size for classification. Correlation filtering was applied to remove highly correlated features, leaving 10 measures. The best classification performance was achieved by combining HNR, RPDE, DFA, and PPE. The results show that PPE alone provides a significant improvement in classification performance, and it appears in all the best performing subsets. The study highlights the limitations of traditional measures in noisy environments and the advantages of non-standard measures in robustness and performance. The findings suggest that non-standard measures are more effective for telemonitoring applications due to their robustness to variations in acoustic environments and individual characteristics. The study also notes that future research should test these findings using voice signals recorded in more typical telemonitoring environments.