Classification of Cough Sounds Using Spectrogram Methods and a Parallel-Stream One-Dimensional Deep Convolutional Neural Network

Abstract

Currently, a subjective method is used to diagnose cough sounds, particularly wet and dry coughs, which can lead to incorrect diagnoses. In this study, novel emergent features were extracted using spectrogram methods and a parallel-stream one-dimensional (1D) deep convolutional neural network (DCNN) to classify cough sounds. The data of this study were obtained from two datasets. We employed the Mel spectrogram, chromagram constant-Q transform, Mel-frequency cepstral coefficient, constant-Q cepstral coefficient, and linear predictive code coefficient to conduct features analysis. The maximum, mean, variance, and standard deviation values of the original spectrogram as well as the maximum first and second derivatives of this spectrogram were extracted and fused to create a single-feature vector. We adopted two types of features: single features and combined features. Each design was restructured according to the magnitude of features with high discrimination power. A parallel-stream 1D-DCNN was developed for classifying cough sounds accurately. We compared the results obtained using the aforementioned network with those obtained using a single-stream 1D-DCNN. We found that the parallel-stream network outperformed the single-stream network for some feature sets. The developed network achieved F1 scores of 98.61% and 82.96% for the first and second datasets, respectively. The concatenation of layers at the flattening level resulted in an F1 score of 99.30% in dataset one. Moreover, layer merging strategies exhibited a better performance at the second convolutional layer level than at the flattening layer level in many cases.

Voice-Cloning Artificial-Intelligence Speakers Can Also Mimic Human-Specific Vocal Expression

“…Studies have noted the importance of Mean F0 and Mean VTL that characterise 'who is talking' or speaker identity (Lavan, Knight, et al, 2019). The current study contributed to the argument that speaker identity and long-term traits, and short-term states, such as speaker emotions, are intertwined (Belin et al, 2004 C#, D, D#, E, F, F#, G, G#, A, A#, and B from lowest to highest in the Western music scale (Huang & Mushi, 2022). Higher Chroma_cqt suggests the speech sample was closer to the higher note in the music scale.…”

Deep Convolutional Neural Networks for the Classification and Detection of Human Vocal Exclamations of Panic in Subway Systems

“…For example, in [13], a random-padding algorithm was proposed to eliminate temporal differences between environmental sound signals in preprocessing. In [14], a zero-padding system was created for preprocessing. In this system, signals of a shorter-thannormal duration are padded with 0s.…”

“…The audio signals in our dataset are spectrogram and timedomain signals. The spectrogram signals were from Mel and log-Mel spectrograms, whose formulas are presented in [14] and [23], respectively. The time-domain signals, which are denoted SIG-WAVE, comprised blended vocal and noise signals.…”

“…Each spectrogram feature was also concatenated with its first and second derivatives. We adopted horizontal [5] rather than vertical concatenation [24], [25], [26], [27] and used a method proposed in our previous study [14], in which multiple image features are combined. The concatenations of Mel and log-Mel spectrogram features with their derivatives are denoted as MEL and log-MEL, respectively.…”

Voice Assistant for Finding Items in Clothing Store

“…The model is trained on 5 different genres and recognizes them with an accuracy of 90.32%. Richard M. used convolutional neural networks to classify cough sounds using features such as mel spectrograms and melfrequency coefficients and achieved an accuracy of 82.96% [5]. Satish K. also conducted research in the field of medicine and used deep learning to monitor the pulse status of patients and achieved 85% accuracy [6].…”