EU program creates criminal network investigation platform with Phonexia voice biometrics

European Union research project ‘ROXANNE’ has completed its development of a system that includes voice biometrics for law enforcement agencies to use in investigating criminal networks.

The project was launched at the Idiap Research Institute in Switzerland in 2019, with the aim of providing criminal network analysis capabilities to law enforcement agencies. In the absence of integrated, automated tools, law enforcement is unable to make full use of the enormous volume of data in extracting evidence and generating actionable intelligence, according to a video presenting the results of the project.

Surveys filled out in 40 countries showed that the volume of data to be processed is the main challenge facing law enforcement trying to break down criminal networks.

The three-year project was funded by the EU Horizon 2020 innovation program. The name is an awkward acronym for “real time network, text, and speaker analytics for combating organized crime.”

The Autocrime platform integrates voice recognition for speaker identification with multilingual automatic speech recognition, gender identification, keyword and topic detection, named entity recognition, and cross-reference and network analysis.

“Ethical and data protection perspectives” are utilized, the video says.

The platform is scalable and flexible, the consortium says, and will be offered to law enforcement in Europe under an open-source license.

The consortium also produced a dataset of synthetic but highly realistic communication for a fictional organized crime network, called ROXSD. The dataset will be made available to researchers in the field of crime fighting and counter-terrorism.

The multidisciplinary consortium consisted of 25 partners from 15 countries, and included participation by 11 law enforcement agencies. Czechia-based Phonexia is the voice biometrics provider in the consortium.

A blog post on the project website explains the role of biometric speaker recognition in the platform. The identities of speakers recorded in criminal investigations are often not known, but biometrics can be used to estimate how likely it is that voices captured in different recordings belong to the same person.

One hundred feature vectors per second are obtained during the signal processing stage, and a deep neural network processes each “to produce a new modified vector.”

Three field test events were held by the consortium, based on realistic scenarios, to demonstrate the platform’s capabilities and gather feedback from users. The group says a “strong majority” found the platform useful and believe it can speed up investigations.

Article Topics

biometric identification  |  biometrics  |  biometrics research  |  criminal ID  |  EU  |  Idiap  |  law enforcement  |  Phonexia  |  ROXANNE  |  speaker recognition  |  voice biometrics