Table 1 This table shows various techniques employed in previous relevant audio related researches along with a research focus (purpose) and publish year.
From: Gun identification from gunshot audios for secure public places using transformer learning
Techniques | Purpose | Year | Accuracy (%) |
|---|---|---|---|
Audio classification | 2003–2007 | 60.0–80.4 | |
Multi layer perceptron21 | Audio classification | 2008 | 70.1 |
One-class SVM22 | Audio classification | 2009 | 76.3 |
Neural network23 | Feature extraction | 2010 | 80.0 |
Deep neural network24 | Audio classification | 2013 | 85.2–8 |
Audio Classification | 2014, 2015 | 86.1 | |
LSTM, RNN28 | Audio classification | 2016 | 88.2–89.3 |
Audio tagging, deep feature extraction | 2017 | 90.3–91.5 | |
Deep unsupervised learning, Unsupervised learning, weakly supervised learning, attention network32,33,34,35 | Audio event detection, audio representation, audio classification | 2018 | 89.0–92.0 |
Few-shot attention, graph neural network, adversarial feature, capsule network36,37,38 | Audio classification | 2019 | 89.2–91.5 |
Audio classification | 2020 | 90.2-91.8 | |
Attention-based networks, zero-shot federated learning41 | Audio classification | 2021 | 91.0-92.5 |
Proposed approach | Audio classification | 2022 | 93.8 |
