Acoustic speech analysis for hypernasality detection in children

Area of Science:

• Speech Science
• Biomedical Engineering
• Computational Linguistics

Background:

• Hypernasality, a speech disorder, poses challenges for accurate diagnosis.
• Automatic identification of hypernasality relies on analyzing acoustic features of speech.
• Existing methods may require extensive training data and significant computational resources.

Purpose of the Study:

• To analyze acoustic features for effective automatic identification of hypernasality.
• To develop a specialized diagnostic feature for hypernasal speech.
• To reduce computational requirements and eliminate the need for training samples in hypernasality detection.

Main Methods:

• Preprocessing of acoustic features using statistical independence analysis.
• Synthesis of a novel diagnostic feature by analyzing acoustic emissions in hypernasal speech.
• Evaluation of feature effectiveness in differentiating hypernasality from normal speech.

Main Results:

• Acoustic features demonstrate sufficient precision in differentiating hypernasality.
• The proposed diagnostic feature effectively identifies the pathology.
• The new feature requires less computational power and no training samples.

Conclusions:

• Acoustic feature analysis is a viable method for automatic hypernasality identification.
• The developed diagnostic feature offers an efficient and accessible approach to detecting hypernasality.
• This method has the potential to improve diagnostic tools for speech disorders.