Sound Recognition Aimed towards Hearing Impaired Individuals in Urban Environment using Ensemble Methods
Automated content-based sound classification is an emerging research avenue with applications in sound analysis, surveillance, noise source identification, multimedia retrieval, smart homes/cities, and urban informatics. Traditionally, hearing aid users have been manually changing the instrument settings according to prevailing acoustic conditions. This change activates the appropriate frequency response, compression parameters, noise-cancellation parameters etc. that best fit the situation. Automatic sensing and classification of current acoustic conditions and subsequent automatic switching can relieve the deaf aid user from the annoying task of recognizing the acoustic environment and manual switching. This paper deals with the urban sound classification using an ensemble method. The dataset used for this research is an urban sound dataset containing 27 hours of audio recording with 10 sound classes. For sound classification, individual classifiers verses ensemble methods are applied to the urban sound dataset. According to results, ensemble methods are proved more efficient and robust due to higher recognition rates. The results obtained can be beneficial for designing hearing aids with automatic switching based on automatic sound classification in urban conditions.
Copyright (c) 2018 Journal of Information Communication Technologies and Robotic Applications
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.