Temporally Aligned Segmentation and Clustering (TASC) - A generalized framework for behavior time series analysis

Sep 14, 2026

Uploaded by Franzy

Behavior exhibits a complex spatiotemporal structure composed of discrete sub-behaviors, or motifs. Continuous behavior streams require segmentation and clustering to expose these embedded patterns, yet many existing tools are tailored to specific setups and lack the time scale and precision needed in experimental and clinical studies. This work proposes Temporally Aligned Segmentation and Clustering (TASC), an iterative framework that uses temporal linear alignment to compute distances and align recurring motifs within multidimensional time series. TASC alternates between evaluating temporal neighbors against cluster centroids and selecting non-overlapping segments followed by reclustering. Results on semi-synthetic and real experimental and clinical data show improved segmentation and clustering quality, providing a stronger foundation for subsequent research and extendable use across domains requiring high-precision time series segmentation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

Introduction

Behavior quantification: segmentation and clustering

Challenges in experimental and clinical behavior data

Behavior motifs across time scales

Bar-Ilan University Bar-Ilan University Bar-Ilan University Schneider Children's Medical Center 􀂦 Bar-Ilan University November 16th, 2023 [https://doi.org/10.21203/rs.3.rs-3560906/v1](https://doi.org/10.21203/rs.3.rs-3560906/v1) 􀁱 􀅏 This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License No competing interests reported. Temporally Aligned Segmentation and Clustering (TASC): A generalized framework for behavior time series analysis Ekaterina Zinkovskaia 1, Orel Tahary 1, Yocheved Loewenstern 1, Noa Benaroya-Milshtein2,3 Izhar Bar-Gad 1 1 Gonda Multidisciplinary Brain Research Center, Bar-Ilan University, Ramat Gan, Israel 2 The Matta and Harry Freund Neuropsychiatric Tourette Clinic, Schneider Children’s Medical Center of Israel, Petach Tikva, Israel 3 Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel Abstract Behavior exhibits a complex spatiotemporal structure consisting of discrete sub-behaviors, or motifs. Continuous behavior data requires segmentation and clustering to reveal these embedded motifs. The popularity of automatic behavior quantification is growing, but existing solutions are often tailored to specific needs and are not designed for the time scale and precision required in many experimental and clinical settings. Here we propose a generalized framework with an iterative approach to refine both segmentation and clustering. Temporally Aligned Segmentation and Clustering (TASC) uses temporal linear alignment to compute distances between and align the recurring behavior motifs in a multidimensional time series, enabling precise segmentation and clustering. We introduce an alternating-step process: evaluation of temporal neighbors against current cluster centroids using linear alignment, alternating with selecting the best non-overlapping segments and their subsequent reclustering. The framework is evaluated on semi-synthetic and real-world experimental and clinical data, demonstrating enhanced segmentation and clustering offering a better foundation for consequent research. The framework may be used to extend existing tools in the field of behavior research and may be applied to other domains requiring high precision of time series segmentation. Introduction Neuroscience experiments investigating behavior and its underlying mechanisms, have been traditionally confined to observing and quantifying either constrained movement or limited subsets of it 1, in both experimental animal models and in human subjects. Behavior quantification in those scenarios could be approximated either by using the experimental setup, e.g., the timing between a cue and a lever press can help approximate timing and characteristic of a reaching movement 2, or be meticulously labeled by a trained expert 1,2 . Moreover, it is common to group behaviors over repetitive trials, which allows to investigate behavior and neuronal characteristic of each group 1. However, such methods cannot be applied to the natural behavior experiments due to the lack of the strict setup and trial organization, nor can the manual labeling scale up to complex behavioral or clinical experiments. Recent advancementsin sensor technology, signal analysis, and especially machine learning, have catalyzed a shift towards automated approaches for behavior analysis 3. Behavior quantification forms the basis for analyzing the mechanisms governing normal and pathological behaviors. Two crucial steps towards behavior quantification are segmentation and clustering 4. Segmentation is a process that entails dividing the data embedded within a continuous time series into short periods (segments), each representing a recurring subbehavior, often termed behavior motif (Fig. 1) . The latter refers to assigning such segments into meaningful clusters, each representing a single underlying sub-behavior. The two are often approached separately, but when coupled can enhance bot

FAQs

What problem does TASC address in behavior analysis?

TASC addresses the need to segment and cluster continuous behavior data to reveal recurring motifs with high temporal precision. It overcomes limitations of existing solutions that are often too specific and insufficient for the required time scale and precision in experimental and clinical settings.

On what kinds of data is TASC evaluated, and what is the reported outcome?

TASC is evaluated on both semi-synthetic and real-world experimental and clinical data. The results demonstrate enhanced segmentation and clustering, supporting better downstream research.

Sep 14, 2026

Uploaded by Franzy