TY - UNPB
T1 - Automatic Quality Control and Enhancement for Voice-Based Remote Parkinson’s Disease Detection
AU - Poorjam, Amir Hossein
AU - Kavalekalam, Mathew S.
AU - Shi, Liming
AU - Yordan, Raykov
AU - Jensen, Jesper B.
AU - Little, Max A
AU - Christensen, Mads Græsbøll
N1 - © 2019 The Authors
PY - 2019/5/28
Y1 - 2019/5/28
N2 - The performance of voice-based Parkinson’s disease(PD) detection systems degrades when there is an acoustic mismatchbetween training and operating conditions caused mainlyby degradation in test signals. In this paper, we address thismismatch by considering three types of degradation commonlyencountered in remote voice analysis, namely background noise,reverberation and nonlinear distortion, and investigate howthese degradations influence the performance of a PD detectionsystem. Given that the specific degradation is known, we explorethe effectiveness of a variety of enhancement algorithms incompensating this mismatch and improving the PD detectionaccuracy. Then, we propose two approaches to automaticallycontrol the quality of recordings by identifying the presenceand type of short-term and long-term degradations and protocolviolations in voice signals. Finally, we experiment with usingthe proposed quality control methods to inform the choice ofenhancement algorithm. Experimental results using the voicerecordings of the mPower mobile PD data set under differentdegradation conditions show the effectiveness of the quality controlapproaches in selecting an appropriate enhancement methodand, consequently, in improving the PD detection accuracy. Thisstudy is a step towards the development of a remote PD detectionsystem capable of operating in unseen acoustic environments.
AB - The performance of voice-based Parkinson’s disease(PD) detection systems degrades when there is an acoustic mismatchbetween training and operating conditions caused mainlyby degradation in test signals. In this paper, we address thismismatch by considering three types of degradation commonlyencountered in remote voice analysis, namely background noise,reverberation and nonlinear distortion, and investigate howthese degradations influence the performance of a PD detectionsystem. Given that the specific degradation is known, we explorethe effectiveness of a variety of enhancement algorithms incompensating this mismatch and improving the PD detectionaccuracy. Then, we propose two approaches to automaticallycontrol the quality of recordings by identifying the presenceand type of short-term and long-term degradations and protocolviolations in voice signals. Finally, we experiment with usingthe proposed quality control methods to inform the choice ofenhancement algorithm. Experimental results using the voicerecordings of the mPower mobile PD data set under differentdegradation conditions show the effectiveness of the quality controlapproaches in selecting an appropriate enhancement methodand, consequently, in improving the PD detection accuracy. Thisstudy is a step towards the development of a remote PD detectionsystem capable of operating in unseen acoustic environments.
M3 - Working paper
BT - Automatic Quality Control and Enhancement for Voice-Based Remote Parkinson’s Disease Detection
ER -