A novel method to identify speech impairment in children using pitch features and stacked LSTM recurrent neural network

Key Points

• Speech impairment identification accuracy improves with pitch features processing using LSTM neural networks, reaching over 85%.
• Key evidence shows a significant correlation between pitch variations and speech impairment in children aged 5-10 years.
• Assessment using a stacked LSTM recurrent neural network demonstrates high efficiency in analyzing speech patterns and features.
• These findings may enable earlier detection of speech difficulties, potentially benefiting interventions for children's language development.