Do You want to know when your machine breaks down or when your data center is under attack?

XiVero’s Anomaly Detection system XiTect is able to recognize any kind of anomaly in the sense of something that deviates from what is standard, normal or expected.

Our Unsupervised Deep Machine Learning based system is able to train itself on normal operational states to detect any anomalous behavior as long as there is a way to extract dedicated features like a spectrum for acoustic surveillance.

Why should you care about anomaly detection?

Cutting costs in maintenance and overall operations as well as reducing the risk of failure is the main reason to go for a predictive approach instead of going down the classical preventive road with all its unnecessary investments.

Why Unsupervised Learning?

In machine learning and especially most of the deep learning approaches it is necessary to train on all categories of data, the system should recognize during operations. Especially in the case of anomaly detection that approach of Supervised Learning fails because there are effectively not many anomalous data available to train a neural network to recognize them.

Our Unsupervised approach assures that the training of the Deep Neural Network just works on the fly by observing the normal features of the system under surveillance.
After a short training period, where the Anomaly Detection System saw a number of normal operational states, it is able to detect any deviation from that state much faster than it would be possible within standard maintenance cycles.

Exemplary Areas of Application,

where it would be to expensive to use standard measurement methods that involve highly qualified engineers to judge whether the system operates within or out of normal parameters.


Predictive Maintenance of machines like pumps.


Discovering anomalies in system log files to reveal cyber-attacks.


Visual quality monitoring to detect deficiencies early.


Medical image evaluation to recognize anomalous features.


Financial inconsistency detection like fraudulent transactions.


Discovery of novelties in large amounts of data (Big Data Analytics).

How do we find the best solution for your problem?

We know that our customers got individual problems they want to get solved in an effort and therefore cost minimized manner. This is the reason why we developed a simple procedure to provide you with a tailor-made solution.

Project Setup

  • Establish the way of working
  • Understand the problem

Feature Extraction

  • Identify the features to be extracted from the available data (e.g. Digital Signal Processing on Audio Data to discover faulty parts in rotational machines).
  • Define ways to get the necessary data recorded.

Record and Analyze the Data to apply the Unsupervised Deep Machine Learning

  • Record for example broadband audio data via contact microphones to detect anomalies in machines like faulty bearings.
  • Analyze system logs to reveal anomalous patterns to detect cyber-attacks early.
  • Identify anomalies in medical images.

Push the trained Deep Neural Network into Operations

  • The trained DNN gets deployed either on cost efficient IoT platforms or even within a cloud to monitor your systems under surveillance.

Data Visualization and Alarming

  • Alert the maintenance personal as soon as XiTect detects any deviation from the normal state.
  • Classical signal- and statistical-processing methods support the engineers to identify faulty parts.

Why should you choose the XiVero to find a solution for your individual Problem of Anomaly Detection?

We are not just experts in Deep Machine Learning, but we’re also proficient in the areas of Digital Signal Processing (DSP) and especially Audio Signal Processing, which enables us to unite the domains of signal analysis and therefore feature extraction with the manifold discipline of Artificial Intelligence.