Introduction

The human voice is thought of as one of our most personal characteristics. People tend to converge towards the language they observe around them, whether it’s copying word choices, mirroring sentence structures or mimicking pronunciations. In the Odyssey (Homer, 850 BC), Helen of Troy is said to have circled the wooden horse while yelling to different Greek soldiers in the voices of their wives and sweethearts because she suspected betrayal. This tale is noteworthy for two different reasons. That is one of the first examples of voice mimicry that has been documented. It may also be the earliest instance of vocal mimicry used for deception (Singh et al., 2017).

This research study is primarily concerned with the identification of speakers whose voices have been disguised and focuses on speaker verification and recognition on the basis of the intonation pattern (www.research.csc.ncsu.edu). The intonation model is typically evaluated based solely on the similarities and differences between several samples, without any explicit reference. Mimicry voice is using synthetic speech against speaker verification based on the spectrum and pitch analysis (Hautamäki et al., 2017). Pitch refers to the highness or lowness of the voice and articulation is the way you pronounce individual sounds. While the generation of speech sounds is the emphasis of articulation, whether you pronounce them correctly is the subject of pronunciation. Both impressions of believability and intelligibility are influenced by the sound characteristics of articulation and pronunciation. During person-to-person interactions, your speech frequently has distinct, strong tones. Articulatory phonetics refers to the equipment used to produce speech sounds as well as the cognitive and physical parameters that specify the range of potential speech sounds and sound patterns. The size and shape of the speaker's vocal tract have an impact on the variety of sounds that can be produced by a human. The oral and nasal cavities, the glottis, the tongue, the velum or soft palate, the hard palate, the teeth, and the lips make up the vocal tract (Delvaux et al., 2017). With so many factors influencing human speech, it becomes simple for us to identify a person just by their voice. So, these speech traits effectively don't alter even when someone tries to impersonate another person, making them useful for detecting disguised voices.

Additionally, the person trying to mimic the voice of another person has their own speech traits, which will occasionally come through in the imitation. Another method of verification is carried out using the PRAAT software where we compare the spectrograms of the two speakers and look for words that both speakers use frequently to determine whether or not voice impersonation has been done.

Obviously, the situation can change. A single person or two distinct people cannot pronounce the same word or sentence with the same intonation. An expert impersonator or an artist impersonating someone else will also have some distinctions that can identify disguise (Latorre et al., 2014).

Another type of speech disguise is self-disguise, which involves concealing one's identity to avoid detection when one of one's voice recordings is being examined in court (Hautamäki et al., 2017).

In this paper, we attempt to provide a strategy for avoiding the mimicking voice, which could pose a security problem. Humans have a tendency to mimic the speaking style of some of the famous personalities. But, from a security perspective, using a mimicked voice for any voice recognition system as a stand-in for an existing voice model is a difficult problem. Mimicry voices are highly vulnerable for any speaker recognition system. Voice dialling, banking over a telephone network, database access services, security control for secret information, and remote access to computers are all areas where a mimicking attack is likely to occur. So, to verify the claim speaker, one must choose the speaker's speech file from among the various speaker models already in use in the system (Kanrar & Mandal, 2015).