Phychoacoustic Masking of Delta and Time -Difference Cepstrum Features for Deception Detection

Use of Machine Learning for Deception Detection From Spectral and Cepstral Features of Speech Signals

“…Two questions were designed to get non-deceptive responses and two questions were designed to get deceptive responses. The database used in this research is a set of speech utterances from audio recordings of a male criminal suspect during an interrogation by a law enforcement officer [6], [13]. During the polygraph testing, the criminal was asked a series of questions at three different times of day that resulted in accumulating twelve responses of 'No'.…”

“…Deception detection based on extracted cepstrum features was studied to understand how speech features can be used to detect human emotion and deception. Cepstral representation of an utterance provides a depiction of the local spectral properties of the signal [3], [6], [9]. Each speech utterance was divided into frames within a 20ms duration, with 320 samples, and a 50 percent overlap.…”

“…Deception detection is considerable practical interest in the field of law enforcement and other government agencies to identify the potential deception at the border crossing and in military scenarios for national security applications. It is also used to evaluate reports from informants at embassies and consulates throughout the world [6], [37]. Deception is intentionally causing an individual to accept falsehood as one that is true.…”

“…In recent years, researchers have been studying the multitude of ways in which humans present their deceptive tricks. The research includes speech signal analysis [2], [4], [6], [12]- [15], thermal facial analysis [7], text analysis with BERT [8], and visual cue analysis [9]- [11] which have proven to be viable options to detect deception. The principal difference between the above technologies is that the polygraph measures some human responses like respiratory rate, electro-dermal activity, heart rate, and others through direct contact, which can lead to numerous challenges and complexities in terms of implementation.…”

“…Although these techniques and classification methods are well known separately in the domain of artificial intelligence and signal processing, however, investigating different spectral and cepstral features (i. e. Mel frequency cepstrum coefficients (MFCCs) [6], [37]- [38], [41], spectral roll of point [41], spectral flux [19], [39], [41], spectral centroid [41], spectral compactness, FFT [16], spectral variability, linear prediction cepstrum coefficients [45], logarithmic of energy, fundamental frequency and zerocrossing rate [41], [44]), develop new features (delta cepstrum, time difference cepstrum, delta energy, time difference energy) in conjunction with principal component analysis (PCA), and find an innovative solution at top-level design approach are key novelty of this research work. This research is also aimed to optimize the features and neural network classifiers to achieve a better recognition rate.…”