Using deep learning for acoustic event classification: The case of natural disasters

Rainfall Observation Leveraging Raindrop Sounds Acquired Using Waterproof Enclosure: Exploring Optimal Length of Sounds for Frequency Analysis

This paper proposes a novel method to estimate rainfall intensity by analyzing the sound of raindrops. An innovative device for collecting acoustic data was designed, capable of blocking ambient noise in rainy environments. The device was deployed in real rainfall conditions during both the monsoon season and non-monsoon season to record raindrop sounds. The collected raindrop sounds were divided into 1 s, 10 s, and 1 min intervals, and the performance of rainfall intensity estimation for each segment length was compared. First, the rainfall occurrence was determined based on four extracted frequency domain features (average of dB, frequency-weighted average of dB, standard deviation of dB, and highest frequency), followed by a quantitative estimation of the rainfall intensity for the periods in which rainfall occurred. The results indicated that the best estimation performance was achieved when using 10 s segments, corresponding to the following metrics: accuracy: 0.909, false alarm ratio: 0.099, critical success index: 0.753, precision: 0.901, recall: 0.821, and F1 score: 0.859 for rainfall occurrence classification; and root mean square error: 1.675 mm/h, R2: 0.798, and mean absolute error: 0.493 mm/h for quantitative rainfall intensity estimation. The proposed small and lightweight device is convenient to install and manage and is remarkably cost-effective compared with traditional rainfall observation equipment. Additionally, this compact rainfall acoustic collection device can facilitate the collection of detailed rainfall information over vast areas.

The Role of Radionuclide Imaging in Brain Death Diagnosis

Brain death is defined as the complete and irreversible cessation of the entire brain function, including the brainstem. For the most part, the diagnosis is clinical, and ancillary testing is only needed when clinical criteria are not satisfied. Differences exist in brain death diagnosis policy in the confirmation of brain death with ancillary testing and the particular test used. Demonstration of the absence of cerebral circulation is a reliable indicator of brain death. Currently, there are no agreed-on universal criteria for ancillary imaging investigation. However, several guidelines and meta-analyses have referred to radionuclide imaging as the most reliable, accurate, and validated ancillary imaging procedure in the confirmation of brain death. Whenever available, lipophilic agents should be preferred using tomographic imaging in all or as needed. False results may occur because of slight temporal delays in flow-function interaction, and such findings may carry prognostic information. Detectable cerebral circulation in the clinical presence of brain death most probably indicates that the process of dying is not yet complete. The results of radionuclide studies may also suggest that the loss of viability in a significant proportion of brain tissue is not compatible with life.

Acoustic scene classification based on three-dimensional multi-channel feature-correlated deep learning networks

As an effective approach to perceive environments, acoustic scene classification (ASC) has received considerable attention in the past few years. Generally, ASC is deemed a challenging task due to subtle differences between various classes of environmental sounds. In this paper, we propose a novel approach to perform accurate classification based on the aggregation of spatial-temporal features extracted from a multi-branch three-dimensional (3D) convolution neural network (CNN) model. The novelties of this paper are as follows. First, we form multiple frequency-domain representations of signals by fully utilizing expert knowledge on acoustics and discrete wavelet transformations (DWT). Secondly, we propose a novel 3D CNN architecture featuring residual connections and squeeze-and-excitation attentions (3D-SE-ResNet) to effectively capture both long-term and short-term correlations inherent in environmental sounds. Thirdly, an auxiliary supervised branch based on the chromatogram of the original signal is incorporated in the proposed architecture to alleviate overfitting risks by providing supplementary information to the model. The performance of the proposed multi-input multi-feature 3D-CNN architecture is numerically evaluated on a typical large-scale dataset in the 2019 IEEE AASP Challenge on Detection and Classification of Acoustic Scenes and Events (DCASE 2019) and is shown to obtain noticeable performance gains over the state-of-the-art methods in the literature.

Introduction to the special issue on machine learning in acoustics

The use of machine learning (ML) in acoustics has received much attention in the last decade. ML is unique in that it can be applied to all areas of acoustics. ML has transformative potentials as it can extract statistically based new information about events observed in acoustic data. Acoustic data provide scientific and engineering insight ranging from biology and communications to ocean and Earth science. This special issue included 61 papers, illustrating the very diverse applications of ML in acoustics.