Assessing the performances of different neural network architectures for the detection of screams and shouts in public transportation

Abstract

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

End-to-end environmental sound classification using a 1D convolutional neural network

“…Environmental sound classification is an interesting problem (Sigtia et al, 2016;Stowell et al, 2015) which has different applications ranging from crime detection (Radhakrishnan et al, 2005) to environmental context aware processing (Chu et al, 2009). Moreover, with the increasing interest in smart cities, IOT devices embedding automatic audio classification can be very useful for urban acoustic monitoring (Mydlarz et al, 2017) like intelligent audio-based surveillance system in public transportation (Laffitte et al, 2019).…”

An Initial Machine Learning-Based Victim’s Scream Detection Analysis for Burning Sites

“…Since this application deals with audio files, the implementation of the LSTM network is a good fit for their excellent performance of time series recurrent classification. The research work presented in [53] shows that the LSTM network has outperformed DNN and convolutional neural networks (CNN) in scream detection in noisy environments. This is similar to our case since fire emergency scenes tend to be extremely noisy.…”

Security Detection in Audio Events: A Comparison of Classification Methods

“…As noted before the main problem with other approach in the literature of audio event detection is the data labeling step. In [9] the data were manually cross-labeled by two different audio experts. Authors in [10] used two parallel GMM classifiers to discriminate, respectively, between screams and noise, and between gunshots and noise.…”

“…The units of each layer represent unknown features which explain the data allowing for various level of abstraction and enabling the output layer to discriminate more efficiently the dataset. DNN has been successfully used in the field of automatic speech recognition [9,23].…”

“…used a GMM classifiers trained on several features to detect screams or gunshots [8]; Laffite and al. measured the performances of diverse neural network architectures to detect shouts and scream in transportation means [9]; Valenzise and al. used a short audio frame where the input signal is used to modeling the background sounds [10].…”