A Real-Time Personalized Noise Reduction Smartphone App for Hearing Enhancement
Nasim Alamdari, Shashank Yaraganalu, Nasser Kehtarnavaz · 2018
This paper presents the development of a personalized noise reduction app that is designed to run in real-time with low-latency on smartphone platforms for hearing enhancement purposes. The personalization is achieved by using an unsupervised noise classifier together with a personalized gain adjustment. After applying a Wiener filtering noise reduction, gains in five frequency bands are adjusted by the user to achieve personalized noise reduction depending on the noise environment identified by the classifier. The other signal processing modules of the app include voice activity detection and compression. Publicly available datasets of speech signals and commonly encountered noise signals are used to test the effectiveness of the app. The results obtained by computing a widely used objective speech quality measure indicate the effectiveness of the app for noise reduction.