Effect on emotions on speaker profiling in cases of hate speeches

Introduction:Hate-speech must have some kind of emotional stress, which we can notice in their voice. Due to this emotional stress we can get deception formant frequency pattern in sound spectrogram. In this study, we have evaluated the effect of emotion on voice in case of hate-speech.

Method:For conducting this study, 60 speech samples have been collected virtually. Acoustic Parameters of all 60 speech samples have been analysed through Auditory & Instrument (C SL-4500) Analysis. From this analysis, a Statistic Data has been captured and after that, studying the statistical Data result has been given.

Result:97% of the time negative correlation was observed in flow of speech and 79% of the time negative correlation was observed in dynamic loudness. 100% of the time Strong correlation was observed in pauses, nasality, and pronunciation of the speech. T-score of vowel “A” was observed significant at alpha=0.05 in F₀. While it was not significant at alpha=0.05 in formant frequency and energy. T-score of vowel “E” was observed significant at alpha=0.05 in F! and energy, While it was not significant at alpha=0.05 in F₂ to F₅ and F₀. T-score of vowel “I” was observed significant at alpha=0.05 in F₀ and energy. While it is not significant at alpha=0.05in formant frequency. T-score of vowel “O” was observed significant at alpha=0.05 in F_t and energy. While it is not significant at alpha=0.05 in F₂ to F₅ and F₀.

Conclusion:According to statistic analysis, We conclude that we can identify the speaker even in different emotion state. Because spectrographic parameter dose not show the significant variation in case of hate speech due to emotion. It indicate that its is specific to the speaker.