Reliability of calculating the cepstral peak without linear regression analysis

Area of Science:

• Speech Science
• Acoustic Analysis
• Voice Disorders

Background:

• Cepstral peak prominence (CPP) is a reliable acoustic measure for detecting dysphonia.
• Traditional methods utilize smoothing algorithms and linear regression analysis (Hillenbrand software).
• Recent advancements include cepstral analysis within the Computerized Speech Laboratory (CSL) software.

Purpose of the Study:

• To compare the reliability of dysphonia prediction using CSL cepstral peak values versus Hillenbrand's method.
• To evaluate the impact of linear regression analysis in cepstral peak prominence calculation.

Main Methods:

• Calculation of cepstral peak values using CSL software.
• Calculation of cepstral peak prominence using Hillenbrand's smoothing algorithm and linear regression.
• Comparison of the predictive accuracy of both methods for dysphonia.

Main Results:

• The CSL software calculates cepstral peak by subtracting the peak value from the baseline signal.
• Linear regression analysis was found to be a crucial step in accurately calculating cepstral peak prominence.
• Software lacking linear regression analysis may have limited usefulness in predicting dysphonia.

Conclusions:

• The inclusion of linear regression analysis is essential for accurate cepstral peak prominence calculation.
• The CSL software's method of cepstral peak calculation may be less effective for dysphonia prediction due to the absence of linear regression.
• Further research may be needed to optimize cepstral analysis for clinical voice assessment.