Scientists from the University of Lagos (University of Lagos, Nigeria) have developed a new method for biometric identification of a person. The developers have found out that laughter acts as a distinctive feature unique to each person.
Laughter is defined by scientists as one of the important natural phenomena in social interactions – a person’s reaction to humor or tickling, manifested in involuntary movements of the muscles of the face and respiratory system and specific sounds.
Engineers from the systems engineering department of the University of Lagos have come to the conclusion that this reaction can also be used as a strong biometric password. A study by Nigerian scientists entitled “Laughter signature: a novel biometric trait for person identification” was published in the International Journal of Biometrics. Human intelligence is able to identify people by their laughter, the abstract to the article says, and it is practically impossible to imitate. However, to date, laughter has not been seen as a potential dynamic biometric identification system. The work, the results of which are published in the publication, is devoted to the creation of the foundations of a new behavioral biometrics model based on individual laugh frequencies for each individual.
According to the popular science journal Naked Science, the University of Lagos team applied a statistical analysis of various sound frequencies characteristic of human laughter. For each of the mel-cepstral coefficients (MFCC) used to characterize speech signals, a Kruskal-Wallis analysis was performed. The dynamic average MFCC coefficients were developed based on typical MFCC features for training the system using a Gaussian Mixed Distribution Model (GMM) and a Support Vector Machine (SVM). The test results showed that the accuracy of personal identification using the conventional Gaussian noise model is 65%, and according to the SVM model – 90%. The combination of these algorithms increases the efficiency of both models: for SVM – by almost 3% and for GMM – by slightly more than 5%.
Thus, the researchers conclude, the study has shown that laughter is a viable biometric trait for personal identification, and it can be embedded in various artificial intelligence systems.
However, this method, despite the high level of security, has its drawbacks. The most important is that to access data protected by a laugh-based password, the user will need to laugh absolutely sincerely. Thus, the developers of such a biometric system will have to create either a universal method that causes a reaction in the form of laughter in all people, or focus on developing an individual approach. Thus, experts believe that at least for the foreseeable future, laughter is unlikely to be used in data access systems.