An mHealth System for Monitoring Medication Adherence in Obstructive Respiratory Diseases Using Content Based Audio Classification

Deep CNN Sparse Coding for Real Time Inhaler Sounds Classification

“…Nousias et al [13,27] compared feature selection and classification strategies using Spectrogram, Cepstrogram and MFCC with supervised classifiers, such as Random Forest (RF), ADABoost and Support Vector Machines (SVMs), demonstrating high classification accuracy.…”

“…Also, the annotation of the different actions was subsequently verified and completed by a trained researcher and based on visual inspection of the acquired temporal signal. In total, 360 audio files were recorded with a duration of twelve seconds each [13][14][15].…”

“…Throughout the processing of the audio data, no further quantization on the data took place, except the quantization of the CNN weights into clusters of similar values of the convolutional and fully connected layer. The device for the recordings is presented in Reference [13]. Figure 2 depicts an overview of the processing pipeline.…”

“…In order to train and test the proposed classifier, the audio recordings were segmented into inhaler activation, exhalation, inhalation, and noise (referring to environmental or other sounds) by a human expert using a graphical annotation tool [13]. A user interface visualizes the audio samples while the user selects parts of the audio files and assigns a class.…”

“…Motivated by the aforementioned open issues, this study lies on the same track as several data-driven approaches [13][14][15], presenting a method that recognizes the respiration and drug delivery phases on acoustic signals derived from pMDI usage. The main focus of this work is the investigation of acceleration aspects, namely filter pruning, scalar pruning and vector quantization, applied on convolutional neural networks (CNNs).…”