PU team creates AI tool to detect cloned voices
Chandigarh, Sept. 23 -- A nine-member team from the department of anthropology at Panjab University, led by master's student Pratibha and guided by renowned forensic scientist professor Kewal Krishan, has developed an artificial intelligence (AI) software capable of distinguishing AI-generated (cloned) voices from real human voices.
The team, comprising eight women, trained the software using a dataset of 100 audio samples - 50 real human voices and 50 computer-generated ones.
The project, which was part of Pratibha's master's dissertation, employed a Support Vector Machine (SVM) model - a specialised AI tool designed to capture tone, texture, and unique speech patterns. During testing, the software demonstrated an accuracy rate of 80% in detecting cloned voices.
The software, which has been granted copyright by the Government of India's Copyright Office, aims to address the growing threat posed by synthetic voices in digital communication. It holds significant promise for forensic analysis of phone threats, voice examinations in crime investigations, and other security-related scenarios. It can effectively identify whether a voice belongs to a real person or is AI-generated.
Professor Kewal Krishan, recently named among the top 2% scientists of the world by Stanford University and the highest-ranked professor at Panjab University in 2025, said, "This software addresses an urgent need in modern forensic science, where the rapid spread of AI-generated content challenges crime detection and evidence analysis."
The research, conducted over six months with multiple experiments and analyses, is expected to reduce the workload of forensic scientists and provide a practical solution to the misuse of AI-generated voices in fraud, cybercrime, and misinformation.
Krishan added that while the current model is trained on 100 samples, Pratibha plans to pursue a PhD to expand the dataset and improve the software's accuracy. The university intends to retain intellectual property rights to the model and will explore industry partnerships for broader application, Krishan said....
To read the full article or to get the complete feed from this publication, please
Contact Us.