Identification of Voice Disorders Using Long-Time Features and Support Vector Machine With Different Feature Reduction Methods

A New Approach for Detection of Pathological Voice Disorders with Reduced Parameters

“…It is concluded from the experimental results that multi-class SVMs provide good performance for classification of these voice data From the previous studies, [34][35][36][37][38] that showed that the related voices could be classified into normal / pathological as depending on sounds' characteristic features. On the other hand, their used classifier and methods, in [34] as accuracy rate 98.3%, in [35] as accuracy rate 97.01%, in [36] as accuracy rate 100%, in [37] as accuracy rate 94.26%, and in [38] as accuracy rate 98.23%. However, in this study we not only have increased accuracy rate of correct class for pathological and normal classification, but also are able to classify the related voices as four different classes, which are important for diagnosying speech voice' analysis.…”

Detection of vocal disorders based on phase space parameters and Lyapunov spectrum

“…The laryngeal nervous system coordinates these subsystems to produce a recognizable voice [51]. In normal voices, vocal folds function flexibly since there are no irregular changes in vocal folds closure and the mass of vocal folds.…”

“…It has been demonstrated that feature reduction can be beneficial in voice pathology detection/classification tasks [50][51][52][53]. While in feature selection, a subspace of the original space of features is selected, feature reduction maps the feature space into a new one where the new features have different characteristics.…”

Intelligibility Evaluation of Pathological Speech through Multigranularity Feature Extraction and Optimization

“…There are a large number of studies mainly focused on the accurate measurement of the fundamental parameters of the previous researches, such as fundamental frequency, jitter, shimmer, amplitude perturbation quotient, pitch perturbation quotient, harmonics-to-noise ratio, and normalized noise energy. In this article, the long-time and short-time 430-dimensional acoustic parameters (basic acoustics feature set, BAFS) are extracted according to the previous studies in Table 1 [5]. …”

“…Zhou et al extracted time-frequency domain modulated characteristics to analyze pathological voice; a recognition rate of 68.3% is achieved based on NKI-CCRT corpus [4]. Arjmandi et al extracted some widely used long-time acoustic parameters, such as shim, jitter, and HNR, to develop an automatic pathological voice computerized system [5]. Previous studies indicate that the voice change detection can be carried out by long-term acoustic parameters; each individual voice utterance can be quantified by a single vector.…”