Age Prediction by Speech: A Machine Learning Approach Using a Common Speech Dataset

Many fields, specifically those involving personal identification, rely on voice processing. Fields like emergency centre routing, electronic finance, criminal detection, and anti-kidnapping operations frequently utilize human speech production that is categorized based on the age of the person’s development, while text classification based on language is used to evaluate lexical properties. Gender categorization is also utilized when analyzing stereotypes in different media content and machine-learning models based on human biases. The purpose of this paper was to begin a new pipeline based on audio data to predict age combining machine learning and deep learning. Common Voice was the utilized data set covering all areas of data collection, preparation, feature creation, and model adoption. The examined models were as follows: Gradient Boosting, XGBoost, LightGBM, CatBoost, and Neural Networks, with a total accuracy percentage of 90% obtained utilizing Neural Networks. The chosen model was applied for real-world inference, which included a frequent maintenance and updating strategy for long-term use. This research provides an in-depth analysis of voice data integration in age-prediction and informs possible paths for voice in machine learning systems.