Faked speech detection with zero prior knowledge

• Published in: Discover applied sciences Vol. 6; no. 6; p. 288
• Main Authors: Al Ajmi, Sahar Abdullah, Hayat, Khizar, Al Obaidi, Alaa Mohammed, Kumar, Naresh, Najim AL-Din, Munaf Salim, Magnier, Baptiste
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 22.05.2024; Springer Nature B.V; Springer
• Subjects: Applied and Technical Physics; Artificial neural networks; Audio data; Bandwidths; Chemistry/Food Science; Classification; Classifiers; Computer Science; Datasets; Deep learning; Earth Sciences; Engineering; English language; Environment; Fourier transforms; Human communication; Machine learning; Materials Science; Multimedia; Neural networks; Sound; Spectrum allocation; Speech; Human mimicry detection; Speech impersonation; Audio forgery; Speech mimicry; Audio forensics
• ISSN: 3004-9261; 2523-3963; 3004-9261; 2523-3971
• DOI: 10.1007/s42452-024-05893-3

Audio is one of the most used ways of human communication, but at the same time it can be easily misused to trick people. With the revolution of AI, the related technologies are now accessible to almost everyone, thus making it simple for the criminals to commit crimes and forgeries. In this work, we introduce a neural network method to develop a classifier that will blindly classify an input audio as real or mimicked; the word ‘blindly’ refers to the ability to detect mimicked audio without references or real sources. We propose a deep neural network following a sequential model that comprises three hidden layers, with alternating dense and drop out layers. The proposed model was trained on a set of 26 important features extracted from a large dataset of audios to get a classifier that was tested on the same set of features from different audios. The data was extracted from two raw datasets, especially composed for this work; an all English dataset and a mixed dataset (Arabic plus English). For the purpose of comparison, the audios were also classified through human inspection with the subjects being the native speakers. The ensued results were interesting and exhibited formidable accuracy, as we were able to get at least 94 % correct classification of the test cases, as against the 85 % accuracy in the case of human observers. Article Highlights A neural network method for blindly classifying audio inputs as genuine or mimicked, without prior references or information. Deals a scenario wherein exactly only one speech sample, either fake or real but not both, of the purported speaker is available. On datasets composed of both English and its combination with Arabic speech samples, the method achieves a remarkable accuracy of at least 94% in differentiating between real and spoofed voices, surpassing human observer accuracy of 85%.