Multimodal feature extraction and fusion for semantic mining of soccer video: a survey

Oskouie, Payam; Alipour, Sara; Moghadam, Amir Masoud Eftekhari

doi:10.1007/s10462-012-9332-4

Cited by 33 publications

(10 citation statements)

References 77 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Consequently, companies such as STATS (Chicago, US), Second Spectrum (Los Angeles, US), ChyronHego (New York, US), and Deltatre (Torino, Italy) have all provided video tracking systems to the market which allow positional data to be collected and used for live and post-match analysis, talent scouting, and media enhancement [3]. A comprehensive survey of the state-of-the-art video-based player tracking systems can be found in Manafifard, Ebadi [4], and a survey on football video analysis was provided by Oskouie, Alipour [5]. Worth mentioning in this context is that a variety of different video tracking systems exist, which can be broadly classified according to the number, arrangement, and specification of utilised cameras (single vs multiple, stationary vs dynamic, stereo vs monocular) [4].…”

Section: Introductionmentioning

confidence: 99%

Football-specific validity of TRACAB’s optical video tracking systems

Linke¹,

Link²,

Lames³

2020

PLoS ONE

130

View full text Add to dashboard Cite

The present study aimed to validate and compare the football-specific measurement accuracy of two optical tracking systems engineered by TRACAB. The "Gen4" system consists of two multi-camera units (a stereo pair) in two locations either side of the halfway line, whereas the distributed "Gen5" system combines two stereo pairs on each side of the field as well as two monocular systems behind the goal areas. Data were collected from 20 male football players in two different exercises (a football sport-specific running course and smallsided games) in a professional football stadium. For evaluating the accuracy of the systems, measures were compared against simultaneously recorded measures of a reference system (VICON motion capture system). Statistical analysis uses RMSE for kinematic variables (position, speed and acceleration) and the difference in percentages for performance indicators (e.g. distance covered, peak speed) per run compared to the reference system. Frames in which players were obviously not tracked were excluded. Gen5 had marginally better accuracy (0.08 m RMSE) for position measurements than Gen4 (0.09 m RMSE) compared to the reference. Accuracy difference in instantaneous speed (Gen4: 0.09 m�s-1 RMSE; Gen5: 0.08 m�s-1 RMSE) and acceleration (Gen4: 0.26 m�s-2 RMSE; Gen5: 0.21 m�s-2 RMSE) measurements were significant, but also trivial in terms of the effect size. For total distance travelled, both Gen4 (0.42 ± 0.60%) and Gen5 (0.27 ± 0.35%) showed only trivial deviations compared to the reference. Gen4 showed moderate differences in the low-speed distance travelled category (-19.41 ± 13.24%) and small differences in the high-speed distance travelled category (8.94 ± 9.49%). Differences in peak speed, acceleration and deceleration were trivial (<0.5%) for both Gen4 and Gen5. These findings suggest that Gen5's distributed camera architecture has minor benefits over Gen4's single-view camera architecture in terms of accuracy. We assume that the main benefit of the Gen5 towards Gen4 lies in increased robustness of the tracking when it comes to optical overlapping of players. Since differences towards the reference system were very low, both TRACAB's tracking systems can be considered as valid technologies for football-specific performance analyses in the settings tested as long as players are tracked correctly.

show abstract

Section: Introductionmentioning

confidence: 99%

Football-specific validity of TRACAB’s optical video tracking systems

Linke¹,

Link²,

Lames³

2020

PLoS ONE

130

View full text Add to dashboard Cite

show abstract

“…To enhance the classifier performance, the misclassification error should be decreased during the learning process. So thus, the misclassification error should be minimized according to cost function in formula (7).…”

Section: Generalized Learning Vector Quantization (Glvq)mentioning

confidence: 99%

“…As the data has multiple peaks it needs custom techniques to analyze the data itself [2,3]. The study of multi-modal data analysis is conducted in the various area e.g., facial expression, sentiment analysis, medical image analysis, video analysis, and surveillance [4][5][6][7][8]. The other researches have been conducted in the special case of multi-modal classification such as large-scale and real-time multi-modal classification [9,10].There are several approaches proposed to solve classification in multi-modal data, such as using a more complex architecture classifier with kernel combination, incremental learning, and ensemble learning [11][12][13].…”

Section: Introductionmentioning

confidence: 99%

Clustering versus Incremental Learning Multi-Codebook Fuzzy Neural Network for Multi-Modal Data Classification

et al. 2020

View full text Add to dashboard Cite

One of the challenges in machine learning is a classification in multi-modal data. The problem needs a customized method as the data has a feature that spreads in several areas. This study proposed a multi-codebook fuzzy neural network classifiers using clustering and incremental learning approaches to deal with multi-modal data classification. The clustering methods used are K-Means and GMM clustering. Experiment result, on a synthetic dataset, the proposed method achieved the highest performance with 84.76% accuracy. Whereas on the benchmark dataset, the proposed method has the highest performance with 79.94% accuracy. The proposed method has 24.9% and 4.7% improvements in synthetic and benchmark datasets respectively compared to the original version. The proposed classifier has better accuracy compared to a popular neural network with 10% and 4.7% margin in synthetic and benchmark dataset respectively.

show abstract

“…The survey by Oskouie et al [14] discusses the topic of multimodal feature extraction and fusion for semantic mining of soccer video, using both visual, audio, and text features. The main events of interest for a summary here are goal, penalty, booking, shot on target and offside situation.…”

Section: Related Workmentioning

confidence: 99%

Audio-Visual Classification of Sports Types

Gade

Abou-Zleikha

Christensen

et al. 2015

2015 IEEE International Conference on Computer Vision Workshop (ICCVW)

View full text Add to dashboard Cite

In this work we propose a method for classification of sports types from combined audio and visual features extracted from thermal video. From audio Mel Frequency Cepstral Coefficients (MFCC) are extracted, and PCA are applied to reduce the feature space to 10 dimensions. From the visual modality short trajectories are constructed to represent the motion of players. From these, four motion features are extracted and combined directly with audio features for classification. A k-nearest neighbour classifier is applied for classification of 180 1-minute video sequences from three sports types. Using 10-fold cross validation a correct classification rate of 96.11% is obtained with multimodal features, compared to 86.67% and 90.00% using only visual or audio features, respectively.

show abstract

Multimodal feature extraction and fusion for semantic mining of soccer video: a survey

Cited by 33 publications

References 77 publications

Football-specific validity of TRACAB’s optical video tracking systems

Football-specific validity of TRACAB’s optical video tracking systems

Clustering versus Incremental Learning Multi-Codebook Fuzzy Neural Network for Multi-Modal Data Classification

Audio-Visual Classification of Sports Types

Contact Info

Product

Resources

About