Contact: Jan Sedmidubsky (xsedmid [at]

Try our DEMO ! Use this login: demo | pass

GaitQualityAnalyzer is a system for analyzing the quality of human gait, developed under project No. MUNI/G/1585/2019. The system manages human subjects along with their gait style recorded in different time periods in the form of motion capture data. The main system objective is to provide the functionality for searching for similar movement patterns on the level of gait cycles and assessing the quality of the retrieved patterns. This enables determining whether a subject performs better or worse after some circumstance (e.g., underwent surgery). Even if the system is intended to be used for evaluating the suitability of treatments for patients suffering from cerebral-palsy disease, it can be potentially applied to other scenarios in which the quality of movement patterns needs to be analyzed.

The system is implemented as a client-server architecture with a web-based graphical user interface. The interface primarily allows users to: (1) manage subjects, visits of subjects, and gait recordings associated with visits, (2) assess the quality of gait recordings, (3) search for similar visits based on attribute filters (such as age, tags, or visit date) and similarities of their associated gait recordings with respect to a selected query visit of another subject, and (4) determine whether the most relevant retrieved subjects exhibit an improvement or deterioration of gait style. The system integrates the following two standalone modules running at specified ports, which can be easily replaced by other implementations in the future if needed:

  • Similarity of movement patterns. A deep similarity model used to compare movement patterns on the level of gait cycles is learned using an LSTM-based recurrent neural network model on selected attributes of rotation angles and moment power. The module also contains the scripts for re-training the LSTM model in case new training data are available. The system also integrates a handcrafted similarity model that supports comparing gait cycles based on user-defined attributes, to provide a certain degree of explainability.
  • Quality of movement patterns. The default implementation assesses the gait quality by computing the Euclidean distance on selected attributes of rotation angles of normalized gait cycles between an input subject and subjects defined as a norm (a user-defined set of subjects along with their characteristic gait cycles). There is also an alternative implementation of gait-quality assessment using the Gait Deviation Index (GDI), which can be downloaded and easily integrated into the system.

The system, including both the above-mentioned modules, is implemented as the Docker container, which allows users to run the system easily on any platform. The input data about subjects and their visits with recorded gait cycles can be either manually added to the system, or initially uploaded in batch mode. In both cases, the recorded gait-cycle data have to be provided in the specific CSV file format as specified within the input-data pre-processor and accompanied documentation.


  • GaitQualityAnalyzer as the Docker container: here.
  • Documentation of GaitQualityAnalyzer: here.
  • External input data pre-processor: here.
  • External gait-quality assessment extractor based on GDI: here.


If you use the GaitQualityAnalyzer system, cite the following software publication:

Sedmidubsky, J., Ljutenko, T., Zezula, P.: GaitQualityAnalyzer. Software, 2022. URL: