From Machine Learning Pioneers to AI Innovators

At Precise Biometrics, the power of Artificial Intelligence (AI) is not just a trend, but a daily reality. Stemming from our roots in *Machine Learning, we have constantly been in the forefront of technological advancements. Machine learning is, for instance, a key success factor behind solutions such as our fingerprint matcher and spoof protection. AI relies on machine learning as its fundamental technology whilst *Deep Learning is a subset of machine learning, and considered an advancement of AI.

The main difference between Machine Learning and Deep Learning is that machine learning relies on manually engineered features, while deep learning automatically learns important features from data itself.

So, how are we using AI here at Precise and what improvements and results have we experienced? Our dedicated team of specialists, is actively harnessing the potential of deep learning within our business unit Algo´s core products:

BioMatch – Our image matcher authenticates & identifies fingerprints

Our primary focus has always been fingerprint authentication, which plays a central role in our business. It addresses the crucial question of determining whether two fingerprint images belong to the same finger (genuine) or not (impostor) with exceptional precision. For several years, our fingerprint authentication algorithms have relied on machine learning techniques. In 2021, we introduced our latest and most advanced algorithm called Evo, which leverages deep learning through Convolutional Neural Networks (*CNN). This innovation has led to a significant enhancement in performance, achieving more than a two-fold improvement in the False Rejection Rate (*FRR) while maintaining the same False Acceptance Rate (*FAR).

BioLive – Our Spoof detection solution

Our spoof detection solution effectively prevents attempts by attackers to fabricate fake fingerprints for authentication purposes. Over the years, we have employed machine learning techniques to address this challenge. With our latest iteration, we have incorporated Convolutional Neural Networks (CNN) to differentiate between genuine fingers and spoofs. This upgraded CNN-based version has generated a remarkable improvement, achieving up to a five-fold enhancement in the Spoof Reject Rate (*SRR). This advancement represents a significant leap forward in the capabilities within this area.

We are also exploring other deep learning benefits within areas such as:

Pattern-based Palm recognition

Palm is an interesting biometric identification method and offers several advantages such as contactless operation, which makes it a more hygienic solution suitable for large settings such as hospitals. In addition to hygiene, palm recognition provides enhanced security and privacy.
We have been working on improving the segmentation of the palms *ROI, Region of Interest, in this case basically the important parts of the hand. The key challenge in  accurately getting the palms ROI is how the hand is positioned. Even a small change in how open the hand is, can affect the lines and wrinkles on the palm a lot, and we have explored ways to measure and understand how the hand is positioned in order to improve the accuracy of getting the palms ROI.

Fingerprint Synthesis using deep learning models

Collecting fingerprint data is costly and time-consuming due to the large number of fingerprint images required to develop and optimize fingerprint matching algorithms. Additionally, how data is collected, utilized, and stored is highly regulated by privacy policies. To mitigate this we have developed a solution that can generate artificial fingerprint images that have the same characteristics as the real fingerprint images captured from a real fingerprint sensor.


Machine Learning – an area within AI, which simply put is a way for computers to learn and make predictions without being explicitly programmed.

Deep Learning – is basically like a “computer brain” that learns and predicts things by copying how our own brains work, to find patterns in information and make smart guesses.

CNN – Convolutional Neural Networks, this is a type of artificial neural network used primarily for image recognition and processing, due to its ability to recognize patterns in images.

FRR – False Rejection Rate, is a metric used to evaluate the performance of a biometric system, by measuring the rate at which genuine users are incorrectly rejected, or not recognized by the system.

SRR – Spoof Rejection Rate, is a metric used to evaluate the performance of a biometric system, by measuring the rate at which spoofs are rejected, or not recognized by the system.

ROI – Region of Interest, refers to a specific area or feature of a biometric sample (e.g. the palm), that is of particular interest for analysis or comparison.