Speech Modeling and Robust Estimation for Diagnosis of Parkinson’s Disease

According to the Parkinson?s Foundation, more than 10 million people world- wide suffer from Parkinson?s disease (PD). The common symptoms are tremor, muscle rigidity and slowness of movement. There is no cure available cur- rently, but clinical intervention can help alleviate the symptoms significantly. Recently, it has been found that PD can be detected and telemonitored by voice signals, such as sustained phonation /a/. However, the voiced-based PD detector suffers from severe performance degradation in adverse envi- ronments, such as noise, reverberation and nonlinear distortion, which are common in uncontrolled settings. In this thesis, we focus on deriving speech modeling and robust estima- tion algorithms capable of improving the PD detection accuracy in adverse environments. Robust estimation algorithms using parametric modeling of voice signals are proposed. We present both segment-wise and sample-wise robust pitch tracking algorithms using the harmonic model. The first order Markov chain is used to impose smoothness prior for the pitch. In segment- wise pitch tracking, we have proposed a method to track the pitch, harmonic order and voicing state jointly based on Bayesian tracking framework. In sample-wise pitch tracking, to deal with colored noise, the noise is modeled as time-varying autoregressive process. The proposed algorithms are com- pared with the state-of-the-art pitch estimation algorithms and are evaluated on the Parkinson?s disease database. Apart from extracting pitch informa- tion, we have also looked into the problem of autoregressive moving average (ARMA) modeling of voiced speech and its parameters estimation. Due to the sparse nature of the excitation signal for the voiced speech, both least 1-norm criterion and sparse Bayesian learning are applied to improve the ARMA co- efficients estimation. The proposed ARMA estimation methods are shown to perform better than the least squares based method in terms of spectral dis- tortion. We have also proposed a dictionary-based speech enhancement algo- rithm using non-negative matrix factorization, where the dictionary items for both speech and noise are parameterized by AR coefficients. Finally, we in- vestigated on the performance of a vast number of speech enhancement and dereveberation algorithms for diagnosis of PD with degraded speech signals.