Wdsae-DNDT Based Speech Fluency Disorder Classification

Abstract

In this paper, Weight Decorrelated Stacked Autoencoder-Deep Neural Decision Trees (WDSAE-DNDT), a novel hybrid model is proposed for automating the assessment of children’s speech fluency disorders by discerning their disfluencies. In fluency disorder classification, it is imperative to know how each feature contributes to the disorder classification rather than the diagnosis itself and so the depth modified DNDT acts as the best discriminator since it is interpretable by its very nature. The WDSAE presents DNDT with a high-level latent representation of the disfluent speech. A fusion feature vector was built by combining the prosodic cues from disfluent speech segments combined with the WDSAE-based Bottleneck features. The proposed hybrid model was compared with the performance of the experimented baseline models. Further analysis was carried out to check the impact of tree cut points for each feature and epochs on the accuracy of prediction of the hybrid model. The proposed hybrid model when trained on the fusion feature set has shown appreciable improvement in the area under the Receiver Operating Characteristics (ROC) curve, classification accuracy, Kappa statistical value, and Jaccard similarity index. The WDSAE-DNDT demonstrates high precision than the baseline models in setting clinical benchmark to distinguish subjects with dysphemia from those with Specific Language Impairment.

Disfluency Assessment Using Deep Super Learners

“…Despite being described as a condition, stuttering can occur in anyone's speech and is frequently triggered by stress or anxiousness. The speech disfluency classification models, which incorporate a unique hybrid deep ensemble [5][6][7] for categorizing varied speech disfluencies in the UCLASS [8] and Fluency Bank datasets [9], achieve an accuracy range of 97 to 98.1 %. FluentNet is a discrete model that comprises a residual CNN [10] that learns frame-level representations of the speech spectrum.…”

An Efficient DenseNet for Diabetic Retinopathy Screening

“…Performance metrics of efficient DenseNet model Table15, the proposed efficient DenseNet model performs well in unique feature extraction for accurate classification of the severity levels of DR, and it has enhanced the efficacy of DR screening. Moreover, the computational complexity[24][25][26][27][28][29] has been reduced compared with the baseline models. The metrics such as precision, recall, and F1 score are used to monitor the grading of DR by the efficient DenseNet model as depicted in Table16, along with the trainable parameters in Table17.…”