Gun identification from gunshot audios for secure public places using transformer learning

Identification of the type of gun used is essential in several fields, including forensics, the military, and defense. In this research, one of the powerful deep learning architectures is applied to identify several types of firearms based on their gunshot noises. For the purpose of extracting features from the audio data, the suggested technique makes use of YAMNet, an effective deep learning-based classification model. The Mel spectrograms created from the collected features are used for multi-class audio classification, which makes it possible to identify different types of guns. 1174 audio samples from 12 distinct weapons make up the study's extensive dataset, which offers a varied and representative collection for training and evaluation. We achieve a remarkable accuracy of 94.96% by employing the best hyperparameter changes and optimization methods. The findings of this study make a substantial contribution to the domains of forensics, military, and defense, where precise gun type identification is crucial. Applying deep learning and mel spectrograms to analyze gunshot audio demonstrates itself to be a promising strategy, providing quick and accurate categorization. This research emphasizes the effectiveness and relevance of using YAMNet, an AI-driven model, as a superior answer to the issues of real-world weapon detection.INDEX TERMS Gun type identification,YAMNet,Transfer Learning,Multi-class audio classification.

Section: Related Workmentioning

confidence: 99%

“…Each fold serves as the validation set once. This process is repeated k times, and the average performance over all folds is calculated to get the final accuracy [5] [14].…”

Section: Proposed Modelmentioning

confidence: 99%

Enhancing Gun Detection With Transfer Learning and YAMNet Audio Classification

Valliappan,

Pande,

Reddy Vinta

2024

“…These results were obtained by using no temporal pooling between convolutional layers. This minimizes the input temporal RF of the network down to 7 frames, as illustrated in Figure 8, which means that the CNN output features correspond to input T-F regions with smaller amounts of overlaps 6 . Hence, more temporal resolution is available to SSRP to distinguish salient T-F regions from other less relevant regions.…”

Section: Impact Of Hyper-parametersmentioning

confidence: 99%

“…In recent years, development of Environment Sound Classification (ESC) methods has been one of the hottest topics in audio classification domain due to its potential use in various application areas such as smart cities [1], audio surveillance systems [2], health care [3], security control systems [4], and Internet of Things (IoT) [5]. For example, ESC can be used to automatically identify different environmental sound events, such as gun shots [6], siren [7], and bird sounds [8]. The quite limited knowledge of temporal and frequency characteristics, the lower signal to noise ratio, and less static patterns make ESC more challenging than other audio-related tasks such as music genre classification and speech recognition.…”

Section: Introductionmentioning

confidence: 99%

Environmental Sound Classification With Low-Complexity Convolutional Neural Network Empowered by Sparse Salient Region Pooling

Seresht

Mohammadi

2023

Environmental Sound Classification (ESC) is an important field in a broad range of applications, such as smart cities, audio surveillance, and health care. Recently, Convolutional Neural Networks (CNNs) have taken the lead from traditional approaches and have produced promising results. However, the achieved improvements are often accompanied by increasing depth, complexity, and size of the network, which prevents their usage in many practical applications. In this work, our goal is to empower a small-size low-complexity CNN model to achieve superior performance. To this end, we concentrate on the importance of global pooling technique, which is less investigated in ESC. In most previous works, models utilize global average pooling layer which does not consider regional saliency, and thus weakens the salient time-frequency regions contributions to the classification, and also to the training of convolutional kernels. We propose a novel global pooling method, called Sparse Salient Region Pooling (SSRP), which computes the channel descriptors using a sparse subset of features, and guides the model to effectively learn from the more salient time-frequency regions. Experimental results demonstrate that the proposed model with only 700K parameters yields accuracies of 86.7% on ESC-50 and 94.8% on ESC-10, which are comparable to that of the state-of-the-art methods. Compared to the baseline model, our model achieves absolute improvement of 21.8% in accuracy on ESC-50, with 98% smaller model size. Our visual analyses show that SSRP intensifies the responses of low-energy regions such that they contribute even more than high-energy regions to the classification of specific sound classes.

“…Cybersecurity professionals can use gunfire recognition technology to identify and trace the sound source to prevent future attacks. Furthermore, these technologies can detect gunshot origins in public places and serve as an early warning system for law enforcement agencies [4]. This could help them respond better and quickly to potential threats while ensuring public safety.…”

Section: Introductionmentioning

confidence: 99%

Preventing Crimes Through Gunshots Recognition Using Novel Feature Engineering and Meta-Learning Approach

Raza,

Rustam,

Mallampati

et al. 2023

Gunshot sounds are common in crimes, particularly those involving threats, harassment, or killing. The gunshot sounds in crimes can create fear and panic among victims, often leading to psychological trauma. Gunshot sounds are associated with a significant mortality rate, especially in cases of gun violence. The sound of gunshots can serve as evidence in criminal investigations, allowing law enforcement officials to determine the number of shots fired, the caliber of the gun used, and the distance from which the shots were fired. Efficient gunshot detection is necessary to address the issue of gun violence in society. This study aims to detect gunshot sounds using an efficient approach to prevent crimes. The frequency-time domain spectrum analysis is performed to understand the patterns of signals related to each target class. A novel Discrete Wavelet Transform Random Forest Probabilistic (DWT-RFP) feature engineering approach is proposed, which takes Mel-frequency cepstral coefficients (MFCC) extracted from gunshot sound data as input for feature extraction. A novel meta-learning-based Meta-RF-KN (MRK) is proposed to detect gunshots based on newly created ensemble features with a DWT-RFP approach. For experiments, the gunshot sounds dataset containing 851 audio clips collected from public videos on YouTube from eight kinds of gun models, is used. Advanced machine learning and deep learning techniques are applied in comparison to evaluate the performance of the proposed approach. Extensive experiments show that the proposed MRK approach achieves 99% k-fold accuracy for detecting gunshots and outperforms state-of-the-art approaches. The proposed approach can potentially be used for accurate gunshot detection and to help prevent crimes.