The effect of voice disguise on Automatic Speaker Recognition

Tan, Tiejun

doi:10.1109/cisp.2010.5647131

Cited by 17 publications

(7 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Accent and dialect imitation. In line to the findings of [Zhang and Tan 2008] and [Tan 2010], in which foreign accent was shown to be robust to the FASRS system, some other studies have investigated the vulnerability of speaker recognition systems to dialect and accent disguise. A GMM-based recognition system using 20 MFCC and including delta and acceleration coefficients was used in [Farrús et al 2006a] over a database consisting of several movies excerpts spoken by two different American actors, one of them using British, Irish and Scottish English, apart from his own dialect.…”

Section: Voice Imitation and Modificationmentioning

confidence: 74%

“…Other similar works such as [Zhang and Tan 2008] and [Tan 2010] studied the effect of 10 kinds of voice disguise in a developed system called FASRS (Forensic Automatic Speaker Recognition System), together with normal voices recorded by 20 male students. The analysis showed that the performance of speaker recognition was highly degraded due to voice disguise, differing in several disguising types, except for the foreign accent, to which it was highly resistant -since FASRS was developed as a language and dialect independent system-, being whisper and masking on mouth the ones which had the greatest effect on the system.…”

Section: Voice Imitation and Modificationmentioning

confidence: 99%

See 1 more Smart Citation

Voice Disguise in Automatic Speaker Recognition

Farrús

2018

ACM Comput. Surv.

View full text Add to dashboard Cite

We, humans, are able to identify other people even in voice disguise conditions. However, we are not immune to all voice changes when trying to identifying people from voice. Likewise, automatic speaker recognition systems can also be deceived by voice imitation and other types of disguise. Taking into account the voice disguise classification into the combination of two different categories (deliberate/non-deliberate and electronic/non-electronic), this survey provides a literature review on the influence of voice disguise in the automatic speaker recognition task and the robustness of these systems to such voice changes. Additionally, the survey addresses existing applications dealing with voice disguise and analyses some issues for future research.

show abstract

Section: Voice Imitation and Modificationmentioning

confidence: 74%

Section: Voice Imitation and Modificationmentioning

confidence: 99%

Voice Disguise in Automatic Speaker Recognition

Farrús

2018

ACM Comput. Surv.

View full text Add to dashboard Cite

show abstract

“…For identification of speaker, a lot of identification process is now used for speaker recognition field. This system used a verification tool for verification of an unknown speaker [20]. The speaker's particular information is being generated by the rapid change in a different features of the speech production technique.…”

Section: Methodology Used For Feature Extractionmentioning

confidence: 99%

Statistical Analysis of Lower and Raised Pitch Voice Signal and Its Efficiency Calculation

Singh¹,

Nandan²,

Kumar³

2019

View full text Add to dashboard Cite

The voice signals of human are a type of acoustic signal that transfers the information about the message or word delivered in the form of speech. The speech of each person has its unique acoustic features. The statistical analysis of such features is critical to the speech recognition. Therefore, this paper aims to identify the speaker through statistical analysis of acoustic features of voice signals. First, the data collection method for speech samples was introduced, the voice signals were divided into three categories, namely, normal voice (NV), lower pitch (LP) and raised pitch (RP), and the effects of the LP and RP on speech were discussed. Then, a feature extraction method was coupled with several classifers to identify the LP and RP for speaker identification. Next, the MFCC, ΔMFCC and ΔΔMFC were adopted to extract the acoustic features. Finally, the proposed method was verified through a speaker identification experiment. The results show that our method can accurately capture the acoustic features of each speaker, and correctly differentiate between the NV, LP and RP. The research results are of great significance to speech recognition and speaker identificaiton.

show abstract

“…Research on such detection has not been reported. The related studies include the work by Tan [2] discussing effect of disguised voice on speaker recognition, and the work by Jin et al [5] studying using voice disguise for speaker de-identification. Detection of disguised voice is not considered in these earlier studies, while it is studied in our present work.…”

Section: Relation To Prior Workmentioning

confidence: 99%

“…Several related studies focus on effect of voice disguise on speaker recognition. It is stated in [2], [3] and [4] that disguised voice can degrade Thanks to 973 Program (2011CB302204) and NSFC (U1135001, 61100168) for funding. speaker recognition performance.…”

Section: Introductionmentioning

confidence: 99%

Blind detection of electronic disguised voice

Wang

Huang

2013

2013 IEEE International Conference on Acoustics, Speech and Signal Processing

View full text Add to dashboard Cite

Since voice disguise has great negative impact on establishing authenticity of audio evidence in forensics, and has shown an increasing tendency in illegal applications, it is important to identify whether a suspected voice has been disguised or not. However, research on such detection has not been reported. In this paper, we focus on blind detection of electronic disguised voice. Statistical moments of Melfrequency cepstrum coefficients (MFCC) are extracted as acoustic features of speech signals. Then an approach for detection of disguised voice based on the extracted features and Support Vector Machine (SVM) classifiers is proposed. The extensive experiments demonstrate that detection rates higher than 95% can be achieved, indicating that detection performance of the proposed approach is good.Index Terms-electronic voice disguise, blind detection, MFCC statistical moments, SVM

show abstract

The effect of voice disguise on Automatic Speaker Recognition

Cited by 17 publications

References 6 publications

Voice Disguise in Automatic Speaker Recognition

Voice Disguise in Automatic Speaker Recognition

Statistical Analysis of Lower and Raised Pitch Voice Signal and Its Efficiency Calculation

Blind detection of electronic disguised voice

Contact Info

Product

Resources

About