Cross-Referencing Self-Training Network for Sound Event Detection in Audio Mixtures.

IEEE Trans Multimedia

Department of Electrical and Computer Engineering and jointly with the Department of Psychology and Brain Sciences, Johns Hopkins University, Baltimore, MD, 21218 USA.

Published: May 2022

Sound event detection is an important facet of audio tagging that aims to identify sounds of interest and define both the sound category and time boundaries for each sound event in a continuous recording. With advances in deep neural networks, there has been tremendous improvement in the performance of sound event detection systems, although at the expense of costly data collection and labeling efforts. In fact, current state-of-the-art methods employ supervised training methods that leverage large amounts of data samples and corresponding labels in order to facilitate identification of sound category and time stamps of events. As an alternative, the current study proposes a semi-supervised method for generating pseudo-labels from unsupervised data using a student-teacher scheme that balances self-training and cross-training. Additionally, this paper explores post-processing which extracts sound intervals from network prediction, for further improvement in sound event detection performance. The proposed approach is evaluated on sound event detection task for the DCASE2020 challenge. The results of these methods on both "validation" and "public evaluation" sets of DESED database show significant improvement compared to the state-of-the art systems in semi-supervised learning.

Download full-text PDF

Source
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC10621403PMC
http://dx.doi.org/10.1109/tmm.2022.3178591DOI Listing

Publication Analysis

Top Keywords

sound event
24
event detection
20
sound
9
sound category
8
category time
8
event
6
detection
5
cross-referencing self-training
4
self-training network
4
network sound
4

Similar Publications

Global declines in wild mussel populations and production have been linked to the impacts of climate change and pollution. Summer die-offs of mussels (Perna canaliculus), spat retention issues, and a severe decline in mussel spat settlement have been reported in the Marlborough Sounds, an important area for mussel farming in New Zealand. Preliminary evidence suggests that naturally occurring contaminants and changing land use in the surrounding areas, could contribute to the decline of this species.

View Article and Find Full Text PDF

What Is Grazing Time? Insights from the Acoustic Signature of Goat Jaw Activity in Wooded Landscapes.

Sensors (Basel)

December 2024

Rangeland Service, Ministry of Agriculture and Food Security, P.O. Box 30, Rishon LeZion 5025001, Israel.

Acoustic monitoring facilitates the detailed study of herbivore grazing by generating a timeline of sound bursts associated with jaw movements (JMs) that perform bite or chew actions. The unclassified stream of JM events was used here in an observational study to explore the notion of "grazing time". Working with shepherded goat herds in a wooded landscape, a horn-based acoustic sensor with a vibration-type microphone was deployed on a volunteer animal along each of 12 foraging routes.

View Article and Find Full Text PDF

The aims of this study were to assess the effect of meloxicam at the time of lameness treatment for hoof-horn (HH) lesions in dairy cattle on 1) time to lameness soundness post trimming and block application and 2) reproductive success Five seasonal-calving pasture-based dairy farms located in the Waikato region of New Zealand were enrolled into a randomized clinical interventional trial. Farmers were tasked with identifying lame animals over a period of approximately -4 to 6 weeks relative to the herd start of mating (HSM), with lameness subsequently confirmed with lameness score (LS) by trained technicians (0-3 scale, where ≥2 are considered lame). Animals with a LS ≥2 were examined by a veterinarian who then enrolled animals if they presented lame with HH (white-line or sole) lesions.

View Article and Find Full Text PDF

Objective: Evaluate the effect of white noise intervention on sleep quality and immunological indicators of patients with breast cancer undergoing neoadjuvant chemotherapy (NAC).

Methods: From January 2020 to December 2022, 104 newly diagnosed female patients (the number of people who met the inclusion criteria) with breast cancer who were confirmed to be preoperative NAC by puncture pathology were selected for a randomised single-blind trial. The patients were randomly divided into an observation group and a control group, with 52 cases in each group.

View Article and Find Full Text PDF

Background: Trimming is critical for a functioning equine hoof. Pressure distribution provides information on loading; however, information on the effects of trimming on pressure distribution is lacking.

Objectives: To describe the pressure changes of equine fore feet following trimming.

View Article and Find Full Text PDF

Want AI Summaries of new PubMed Abstracts delivered to your In-box?

Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!