Audio Recognition: a new horizon for public safety and forensic investigation

Audio Recognition software offers great benefits to forensic experts and public safety organizations by helping them identify suspects or criminals through audio recordings.

Voice Biometrics and Audio Recognition

We explained in a previous article on Biometric Voice Recognition that it is a technology capable of identifying a person starting from the biometric characteristics of the voice.

If such technology is enough to recognize a voice, why are we talking about Audio Recognition?

In the forensic and investigative fields, Biometric Voice Recognition can certainly help to confirm or refute the identity of a speaker in audio recordings. But the Artificial Intelligence on which Audio Recognition is based can go even further. In fact, it does not just recognize a voice, it can also identify background noises. These can be important clues to reconnect certain people to specific places.

Speaker Diarization: integration into Audio Recognition software

Another element of great advantage in investigations is offered by the Speaker Diarization. It has the task of identifying, in an audio where there are several voices, who is speaking and at what moment. It was initially a job that was up to voice analysts and investigators. The biggest obstacle was often the poor quality of the recording. In recent years, thanks to the evolution of deep learning, important progress has been made in Speaker Diarization. A great turning point is represented by the integration of this technology into Audio Recognition software.

Audio Recognition: What Benefits for Public Safety?

A study published in “Forensic Linguistics” in 2000 showed how the rate of recognition of voices starting from an audio recording, even among acquaintances, was found to be quite low. There was even a volunteer who could not even recognize his own voice. We also remind you that the major obstacle during the investigations is the low quality and loud environmental noises of the recording. There are times when an audio is discarded for this very reason. It is clear that in these cases data that are unusable if left for human listening alone are being collected.

Artificial Intelligence becomes a fundamental support for investigations, in the context of public safety. In particular, Audio Recognition software can recognize voices and ambient sounds even in low quality recordings. It releases information from data that are often indecipherable by the human ear.

This technology allows forensic experts to save countless hours otherwise spent listening to audio recordings. Moreover, it can analyze multiple data at the same time, managing in a short time to return useful clues to investigations. Thanks to Speaker Diarization, it is also possible to identify and differentiate the voices of different speakers. Investigators are thus free to devote themselves to activities of higher added value, speeding up and improving investigations and consequently offering a more efficient service in the field of public safety.

Peculiarities of Polyphonic: Audio Recognition Software Analysis

Polyphonic is the Audio Recognition solution by Pragma Etimos, an innovative scale-up operating in the Data Intelligence and Green Data sector. The software deals with the collection and analysis of sounds and subsequent classification for the identification of human voices and background noises.

This occurs through the identification of the Audiome: sum of Audiosomes that make up the voice / sound imprint. The latter are the unique identifier that helps to compose the imprint of a sound and is given by the frequencies of the sound itself.

Each Audioma identified by Polyphonic it can be the set of two different types of Audiosomes:

• Human audiosome that recognizes who the voice belongs to, and various characteristics.
• Background noise audiosome that recognizes the environment (Airport, train station, bars, restaurants, construction sites, etc).

By connecting these two types of Audiosomes, we can identify a person from the background noise with which they have been classified.

In conclusion Polyphonic offers services of:

• Voice fingerprint identification
• Speaker Diarization
• Identification of the language
• Gender identification
• Estimate of age
• Voice activity detection
• Estimation of speech quality

MORE TO EXPLORE …

Public Safety & Audio Recognition