Machine Learning Based Video Category Classification Using Hybrid Wavelet Transform of Cosine, Haar and Walsh
| Autor: | Hansa M Shimpi, Sudeep D. Thepade |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Learning classifier system
Computer science business.industry Feature vector Frame (networking) ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Decision tree Wavelet transform Pattern recognition Video processing k-nearest neighbors algorithm Support vector machine ComputingMethodologies_PATTERNRECOGNITION Artificial intelligence business |
| Zdroj: | 2018 Fourth International Conference on Computing Communication Control and Automation (ICCUBEA). |
| DOI: | 10.1109/iccubea.2018.8697534 |
| Popis: | In the video processing, video classification has been a fast growing, Challenging and interesting area. With such flourishing of Internet technology and advancement in repository mechanism, the rate of generation of video per second is very large. To manually classify such a voluminous and complex video data becomes a tedious task. This paper investigates the novel content-based video classification with the help of various transform like Cosine, Haar, Walsh and Hybrid Wavelet Transform. Content-based itself define that it purely consider the actual content of video rather than considering user defined text and keywords to be used as signature (feature vector) for representation of entire video in classification. Instead of considered the complete transformed data, only high energy transformed content of video frame is used as feature vector in proposed video classification methods. Distinctive machine learning classifier like K Nearest Neighbor(KNN), Support Vector Machine (SVM) and Decision Tree are used. The fundamental point of grouping is to improve the accuracy. The experimentation have shown that the fractional transformed content gives better accuracy in Cosine, Haar, Walsh and Hybrid Wavelet Transform. The classification accuracy of proposed video classification with hybridization of transform has proven to be better. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|