Related Experiment Videos
Adaptive noise energy estimation in pathological speech signals
1Department of Electronics and Telecommunications, University of Florence, Italy. manfredi@det.unifi.it
IEEE Transactions on Bio-Medical Engineering
|November 15, 2000
Summary
Spectral noise in pathological voices correlates with hoarseness. This study tracks noise variations using an adaptive normalized noise energy method, improving accuracy by adjusting analysis windows based on pitch estimation for clearer hoarseness assessment.
Area of Science:
- Speech pathology
- Acoustic analysis
- Voice disorders
Background:
- Spectral noise is a key acoustic correlate of perceived hoarseness in pathological voices.
- Quantifying noise variations accurately is crucial for understanding voice pathology.
- Existing methods may be limited by fixed analysis parameters.
Purpose of the Study:
- To track spectral noise variations during speech in pathological voices.
- To introduce an adaptive method for noise analysis that accounts for vocal fold dynamics.
- To evaluate the proposed method on synthetic and real patient data.
Main Methods:
- Utilized an adaptive version of the normalized noise energy (NNE) method.
- Incorporated pitch estimation to dynamically adjust the time window length for noise analysis.
- Tested the approach on synthetic voice data and applied it to recordings from cordectomized adult males.
Main Results:
- The adaptive NNE method successfully tracked noise variations during utterances.
- Pitch estimation enabled optimal window selection, avoiding inaccuracies from "dip" regions.
- The method demonstrated applicability to real-world pathological voice data.
Conclusions:
- The adaptive NNE method provides a robust approach for analyzing spectral noise in pathological voices.
- Dynamic window adjustment based on pitch estimation enhances the accuracy of hoarseness-related noise measurement.
- This technique holds potential for improved diagnosis and monitoring of voice disorders.