Browsing by Author "Wang, Jun"
Now showing 1 - 7 of 7
- Results Per Page
- Sort Options
Item Automatic Speech Activity Recognition from MEG Signals Using Seq2Seq Learning(IEEE Computer Society, 2019-03) Dash, Debadatta; Ferrari, P.; Malik, S.; Wang, Jun; 0000-0001-7265-217X (Wang, J); Dash, Debadatta; Wang, JunAccurate interpretation of speech activity from brain signals is critical for understanding the relationship between neural patterns and speech production. Current research on speech activity recognition from the brain activity heavily relies on the region of interest (ROI) based functional connectivity analysis or source separation strategies to map the activity as a spatial localization of a brain function. Albeit effective, these methods require prior knowledge of the brain and expensive computational effort. In this study, we investigated automatic speech activity recognition from brain signals using machine learning. Neural signals of four subjects during four stages of a speech task (i.e., rest, perception, preparation, and production) were recorded using magnetoencephalography (MEG), which has an excellent temporal and spatial resolution. First, a deep neural network (DNN) was used to classify the four whole tasks from the MEG signals. Further, we trained a sequence to sequence (Seq2Seq) long short-term memory-recurrent neural network (LSTM-RNN) for continuous (sample by sample) prediction of the speech stages/tasks by leveraging its sequential pattern learning paradigm. Experimental results indicate the effectiveness of both DNN and LSTM-RNN for automatic speech activity recognition from MEG signals. © 2019 IEEE.Item Detecting Functional Connectivity Changes in Neuroimages of Mesial Temporal Lobe Epilepsy and Type 2 Diabetes Using Machine Learning(2017-05) Famili, Afarin; Montillo, Albert; Maldjian, Joseph; Wang, JunThe purpose of this thesis is to detect the changes in functional connectivity that are detectable in neuroimaging data obtained for mesial temporal lobe epilepsy and type 2 diabetes using machine learning as a tool to uncover and characterize the changes. Thus this thesis is divided into two parts. In the first part, our experiments focus on the study of patients with refractory mesial temporal lobe epilepsy (MTLE). MTLE is known to cause memory deficits and the preferred treatment to suppress seizures, temporal lobe resection, can exacerbate difficulties in forming new episodic memory. Therefore, learning about the success or failure to form new memory in such patients is critical if we are to facilitate future development of new therapies based in cognitive biofeedback training or electrical neurostimulation-based therapies. To this end, we address many challenges in analyzing memory formation when MTLE brains are recorded using stereo-encephalography (sEEG) in a Free Recall test of episodic memory. Our contributions are threefold. First, we compute a rich measure of brain connectivity by computing the phase locking value statistic (synchrony) between pairs of regions, over hundreds of word memorization trials. Second, we leverage the rich information (over 400 values per pair of probed brain regions) to form consistent length feature vectors for classifier training. Third, we train and evaluate seven different types of classifier models and identify which ones achieve the highest accuracy and which brain features are most important for high accuracy. In the second part, our experiments focus determining whether an association between Type 2 Diabetes (T2D) management and functional brain health in African American (AA) population exists, and if so quantifying that association. In these experiments we assess resting state functional magnetic resonance imaging (rs-fMRI) from adult African Americans with T2D. To form brain health measures, we construct a graph representation of brain connectivity and compute multiple local and global quantitative measures suitable for statistical analysis using graph theory. We then fit a regularized, linear support vector regression model to predict the brain health measure using 10-fold nested cross-validation and apply permutation testing to measure the statistical reliability of the fitted model. Our data supports an association between T2D management and brain health in African Americans. Moreover, we show details of how T2D impacts specific regions of the brain and its functional connectivity. Our findings demonstrate alterations in functional connectivity including the degree of the right hippocampus and degree of the left parahippocampal gyrus with diabetes and cardiovascular disease measures. The findings corroborate the association between CAC and hippocampal gray matter volume found in African Americans previously. This work gives additional impetus to T2 diabetics to control their glycemic load as the impact of diabetes may not be just on cardiac and kidney health but may extend to functional brain health as well.Item Parkinson's Condition Estimation Using Speech Acoustic and Inversely Mapped Articulatory Data(International Speech and Communication Association) Hahm, Seongjun; Wang, Jun; Noth E.; Steidl S.; Moller S.; Ney H.; Mobius B.; Hahm, Seongjun; Wang, JunParkinson's disease is a neurological disorder that affects patient's motor function including speech articulation. There is no cure for Parkinson's disease. Speech and motor function declines as the disease progresses. Automatic assessment of the disease condition may advance the treatment of Parkinson's disease with objective, inexpensive measures. Speech acoustics, which can be easily obtained from patients, has been used for automatic assessment. The use of information in motor function of articulator (e.g., jaw, tongue, or lips) has rarely been investigated. In this paper, we proposed an approach of automatic assessment of Parkinson's condition using both acoustic data and acoustically-inverted articulatory data. The quasi-articulatory features were obtained from the Parkinson's acoustic speech data using acoustic-to-articulatory inverse mapping. Support vector regression (SVR) and deep neural network (DNN) regression were used in the experiment. Results indicated adding articulatory data to acoustic data can improve the performance of using acoustic data only, for both SVR and DNN. In addition, deep neural network outperformed support vector regression on the same data features measured with Pearson correlation but not with Spearman correlation. The implications of our approach with further improvement were discussed.Item Phonological Organization in the Production of Novel Words in Typical and Atypical Language Development(2021-12-01T06:00:00.000Z) Benham, Sara; Goffman, Lisa; Assmann, Peter; Dollaghan, Christine; Katz, William; Wang, JunA core deficit in preschoolers with developmental language disorder (DLD) is in learning words. Words incorporate both sound patterns and the meanings to which they are mapped. Children with DLD show particular deficits in word form rather semantic aspects of word learning (Dollaghan, 1987; Leonard et al., 2019). Deficits are characterized by increased sound errors relative to peers with typical language development (TD); however, little is known about the stability of these errors over multiple productions of the same word. The objective of this project was to investigate two aspects of phonological word form learning—the variability of syllable sequences and the systematicity of sound feature organization—in 4-5-year-olds with DLD, agematched TD peers, and 2-year-olds with typical language. It was predicted that both children with DLD and younger toddlers would show high levels of phonological variability (e.g., Ferguson & Farwell, 1975). In contrast, toddlers, but not children with DLD, were expected to show systematicity in their errors (e.g., Vihman, 1996). A major aim was to determine whether sound pattern deficits attested in DLD represent a qualitative deficit or are developmental in nature. Toddlers are a critical control group for positioning the findings from children with DLD because it is hypothesized that changes in vocabulary size influence both the variability and systematicity of sound patterns. Therefore, this project also incorporates manipulations to assess the lexical and phonological interface. This dissertation includes comprehensive summaries of three manuscripts and two new studies on phonological organization related to syllable sequence variability and sound feature systematicity in two cohorts of children. Cohort 1 consists of 21 preschoolers with DLD and 21 peers with TD. Cohort 2 consists of 13 two-year-olds with TD and 16 four-year-olds with TD. All children participated in a word learning task in which they imitated multiple repetitions of novel word forms with and without a meaningful object referent. Cohort 2 also produced word forms with a relatively simple phonological structure. Comprehension probes were included to assess learning. Based on broad phonetic transcriptions of each production, segmental accuracy and sound feature accuracy of each form were computed. In addition, novel methods based on network science were developed to assess the variability and systematicity of phonological elements at two different levels: the sequential organization of syllables, and the bundling of phonetic features within syllables. Findings from this project reveal that both TD toddlers and preschoolers with DLD show relatively high levels of syllable sequence variability when compared with TD preschoolers, suggesting that sequential errors in the nonword productions of children with DLD are developmental in origin. The variability in sound organization is also not explained by differences in sound feature systematicity, which was similar across all groups studied. That is, toddlers, preschoolers with DLD, and preschoolers with typical language showed similar patterns of sound feature bundling within syllables. Toddlers showed higher variability than TD preschoolers in words with both complex and relatively simpler structure, suggesting that variability is a core feature of their novel word productions even with varying phonological structure. Further, the variability and systematicity of toddlers’ productions were unaffected by the inclusion of a meaningful referent; however, children with DLD showed increases in the stability of syllable sequences and of sound feature organization when a meaningful referent was added. Thus, children with DLD, but not toddlers, showed interactivity across semantic and phonological levels. These findings suggest a similar developmental trajectory for children with DLD and younger toddlers in both the variability and systematicity of phonological form. In contrast, children with DLD are affected by the inclusion of a referent while toddlers are not, suggesting that this aspect of lexical and phonological interaction is not developmental.Item Predicting Speech Intelligibility Decline in Amyotrophic Lateral Sclerosis Based on the Deterioration of Individual Speech Subsystems(Public Library of Science) Rong, Panying; Yunusova, Yana; Wang, Jun; Zinman, Lorne; Pattee, Gary L.; Berry, James D.; Perry, Bridget; Green, Jordan R.; 0000-0001-7265-217X (Wang, J); Wang, JunPurpose: To determine the mechanisms of speech intelligibility impairment due to neurologic impairments, intelligibility decline was modeled as a function of co-occurring changes in the articulatory, resonatory, phonatory, and respiratory subsystems.; Method: Sixty-six individuals diagnosed with amyotrophic lateral sclerosis (ALS) were studied longitudinally. The disease-related changes in articulatory, resonatory, phonatory, and respiratory subsystems were quantified using multiple instrumental measures, which were subjected to a principal component analysis and mixed effects models to derive a set of speech subsystem predictors. A stepwise approach was used to select the best set of subsystem predictors to model the overall decline in intelligibility.; Results: Intelligibility was modeled as a function of five predictors that corresponded to velocities of lip and jaw movements (articulatory), number of syllable repetitions in the alternating motion rate task (articulatory), nasal airflow (resonatory), maximum fundamental frequency (phonatory), and speech pauses (respiratory). The model accounted for 95.6% of the variance in intelligibility, among which the articulatory predictors showed the most substantial independent contribution (57.7%).; Conclusion: Articulatory impairments characterized by reduced velocities of lip and jaw movements and resonatory impairments characterized by increased nasal airflow served as the subsystem predictors of the longitudinal decline of speech intelligibility in ALS. Declines in maximum performance tasks such as the alternating motion rate preceded declines in intelligibility, thus serving as early predictors of bulbar dysfunction. Following the rapid decline in speech intelligibility, a precipitous decline in maximum performance tasks subsequently occurred.;Item “SapSense”- Development of a Field Based Biosensor towards Wheat Pathogen Detection(2017-12) Vasudevan, Akshay; Prasad, Shalini; Nagaraj, Vinay J; Wang, JunWheat is cultivated in almost all the states in the United States of America, and it is the third most cultivated crop in terms of acreage, producing a yield of 60 million tons per year. In 2016, world production of wheat was 749 million tons, making it the second most-produced cereal after maize. The most common viruses that affect wheat in the United States are Wheat Spindle Streak Mosaic Virus (WSSMV), Soil-borne Wheat Mosaic Virus (SBWMV/SBV) and the Yellow Dwarf Virus (YDF)/ High Plains Virus (HPV). Viral infections in wheat often go unnoticed due to several reasons, and cause a loss in yield and consequently a loss in revenue which is estimated at an annual average of $35 million per year, in the United States. Traditional techniques to detect these pathogens are ELISA, dot blots and PCR, all of which are time consuming and/or require benchtop laboratory equipment, and are not suitable for rapid screening. This work aims to design and demonstrate proof of feasibility in designing a biosensor system for screening wheat viruses, through the development of a point of use system which uses the principles of electrochemical impedance spectroscopy to detect the presence of the infection causing virus. The novelty in the proposed work is to be able to detect the virus in a standard buffer, plant sap, removing the need for any kind of filtering and processing, and thereby having the ability to detect the virus in real-time . The biosensor works on the principles of affinity based bio-sensing with an immunoassay built on the surface of a gold electrode. Virus detection is achieved by characterizing impedance changes on the sensor surface associated with the binding of the virus to its affinity antibody probe, which is then measured by electrochemical impedance spectroscopy (EIS).Item Using Electromagnetic Articulography with a Tongue Lateral Sensor to Discriminate Manner of Articulation(Acoustical Society of America) Katz, William F.; Mehta, Sonya; Wood, Matthew; Wang, Jun; Katz, William F.; Mehta, Sonya; Wood, Matthew; Wang, JunThis study examined the contributions of the tongue tip (TT), tongue body (TB), and tongue lateral (TL) sensors in the electromagnetic articulography (EMA) measurement of American English alveolar consonants. Thirteen adults produced /ɹ/, /l/, /z/ and /d/ in /αCα/ syllables while being recorded with an EMA system. According to statistical analysis of sensor movement and the results of a machine classification experiment, the TT sensor contributed most to consonant differences, followed by TB. The TL sensor played a complementary role, particularly for distinguishing /z/. © 2017 Acoustical Society of America.