Há um crescente interesse em sistemas de detecção de violência de forma automática por meio do áudio ambiente. Neste trabalho, construímos e avaliamos 4 classificadores com essa proposta. Porém, em vez de processar diretamente os sinais de áudio, nós os convertemos para imagens, conhecidas como mel-spectrograms, e em seguida utilizamos Redes Neurais Convolucionais (CNN) para tratar como um problema de classificação de imagens utilizando-se de redes pre-treinadas neste contexto. Testou-se as arquiteturas Inception v3, VGG-16, MobileNet v2 e ResNet152 v2, tendo o classificador oriundo da arquitetura MobileNet obtido os melhores resultados de classificação, quando avaliado no HEAR Dataset, criado para a realização desta pesquisa.

According to the World Health Organization, an estimated one billion people live with a disability. Millions of them are non-verbal, and also experience motor-skill challenges. The limitations on activity and restrictions on participation due to such disabilities often lead to discrimination and social exclusion. A UNICEF study analyzing data from 15 countries found that almost 50% of children with disabilities are out of school, and 85% of them did not receive any formal education. Affording enhanced and accelerated communication for disabled people, who continue to form the worlds largest minority to experience social discrimination, is central to making the world a more inclusive place. Our LIVOX application incorporates artificial intelligence algorithms to reduce the so-called reciprocity gap that acts as a communication barrier between disabled people and their interlocutors, thus enabling people with disabilities, especially children, to participate in daily social and educational activities. Integrating them into the existing social structures is central to making the world a more inclusive place.