Multiaxial Haar-Like Feature and Compact Cascaded Classifier for Versatile Recognition

Multiaxial Haar-Like Feature and Compact Cascaded Classifier for Versatile Recognition. Nishimura, J. & Kuroda, T. Sensors Journal, IEEE, 10(11):1786-1795, Nov, 2010.
doi abstract bibtex

A versatile recognition algorithm has been proposed to process image, sound, and 3-D acceleration signals with a common framework at low calculation cost. Firstly, a novel 1-D Haar-like feature is used to roughly extract frequency information from temporal signals. Biaxial and mean-embedded Haar-like features are proposed to extract the standard deviation and the interaxial correlation from 3-D acceleration signals. Secondly, two techniques are proposed to build a compact cascaded classifier. Redundant feature selection (RFS) incorporates the features which are already selected in previous stage classifiers to reduce the calculation cost. A dynamic look-up table (DLUT) is proposed to construct a look-up table-based weak classifier with the smallest possible number of bins. A train loss function is by globally optimized using dynamic programming. The proposed algorithm is tested experimentally on speech/nonspeech classification and human activity recognition. The proposed algorithm yields a speech/nonspeech classification performance comparable to the state-of-art method called MFCC while reducing the calculation cost by 100 times. The algorithm also achieves human activity recognition accuracy of 96.1% with calculation cost reduction of 84% compared with the state-of-art method based on C4.5 decision-tree classifier using the basic statistical features. The proposed algorithm has been employed to build the versatile recognition processor.

@Article{Nishimura2010,
  Title                    = {Multiaxial Haar-Like Feature and Compact Cascaded Classifier for Versatile Recognition},
  Author                   = {Nishimura, J. and Kuroda, T.},
  Journal                  = {Sensors Journal, IEEE},
  Year                     = {2010},

  Month                    = {Nov},
  Number                   = {11},
  Pages                    = {1786-1795},
  Volume                   = {10},

  Abstract                 = {A versatile recognition algorithm has been proposed to process image, sound, and 3-D acceleration signals with a common framework at low calculation cost. Firstly, a novel 1-D Haar-like feature is used to roughly extract frequency information from temporal signals. Biaxial and mean-embedded Haar-like features are proposed to extract the standard deviation and the interaxial correlation from 3-D acceleration signals. Secondly, two techniques are proposed to build a compact cascaded classifier. Redundant feature selection (RFS) incorporates the features which are already selected in previous stage classifiers to reduce the calculation cost. A dynamic look-up table (DLUT) is proposed to construct a look-up table-based weak classifier with the smallest possible number of bins. A train loss function is by globally optimized using dynamic programming. The proposed algorithm is tested experimentally on speech/nonspeech classification and human activity recognition. The proposed algorithm yields a speech/nonspeech classification performance comparable to the state-of-art method called MFCC while reducing the calculation cost by 100 times. The algorithm also achieves human activity recognition accuracy of 96.1% with calculation cost reduction of 84% compared with the state-of-art method based on C4.5 decision-tree classifier using the basic statistical features. The proposed algorithm has been employed to build the versatile recognition processor.},
  Doi                      = {10.1109/JSEN.2010.2049740},
  ISSN                     = {1530-437X},
  Keywords                 = {image classification;image recognition;ubiquitous computing;3-D acceleration signals;MFCC;compact cascaded classifier;dynamic look-up table;image processing;multiaxial Haar-like feature classifier;pervasive computing;redundant feature selection;speech classification;ubiquitous computing;versatile recognition algorithm;Acceleration;Costs;Data mining;Dynamic programming;Frequency;Humans;Image recognition;Signal processing;Speech;Table lookup;Cascaded classifier;Haar-like feature;versatile recognition},
  Timestamp                = {2014.12.22}
}

Downloads: 0

{"_id":"HzidoCRDJj8rJZ8zp","bibbaseid":"nishimura-kuroda-multiaxialhaarlikefeatureandcompactcascadedclassifierforversatilerecognition-2010","downloads":0,"creationDate":"2017-09-14T16:34:36.879Z","title":"Multiaxial Haar-Like Feature and Compact Cascaded Classifier for Versatile Recognition","author_short":["Nishimura, J.","Kuroda, T."],"year":2010,"bibtype":"article","biburl":"https://raw.githubusercontent.com/jfslin/jfslin.github.io/master/jf2lin.bib","bibdata":{"bibtype":"article","type":"article","title":"Multiaxial Haar-Like Feature and Compact Cascaded Classifier for Versatile Recognition","author":[{"propositions":[],"lastnames":["Nishimura"],"firstnames":["J."],"suffixes":[]},{"propositions":[],"lastnames":["Kuroda"],"firstnames":["T."],"suffixes":[]}],"journal":"Sensors Journal, IEEE","year":"2010","month":"Nov","number":"11","pages":"1786-1795","volume":"10","abstract":"A versatile recognition algorithm has been proposed to process image, sound, and 3-D acceleration signals with a common framework at low calculation cost. Firstly, a novel 1-D Haar-like feature is used to roughly extract frequency information from temporal signals. Biaxial and mean-embedded Haar-like features are proposed to extract the standard deviation and the interaxial correlation from 3-D acceleration signals. Secondly, two techniques are proposed to build a compact cascaded classifier. Redundant feature selection (RFS) incorporates the features which are already selected in previous stage classifiers to reduce the calculation cost. A dynamic look-up table (DLUT) is proposed to construct a look-up table-based weak classifier with the smallest possible number of bins. A train loss function is by globally optimized using dynamic programming. The proposed algorithm is tested experimentally on speech/nonspeech classification and human activity recognition. The proposed algorithm yields a speech/nonspeech classification performance comparable to the state-of-art method called MFCC while reducing the calculation cost by 100 times. The algorithm also achieves human activity recognition accuracy of 96.1% with calculation cost reduction of 84% compared with the state-of-art method based on C4.5 decision-tree classifier using the basic statistical features. The proposed algorithm has been employed to build the versatile recognition processor.","doi":"10.1109/JSEN.2010.2049740","issn":"1530-437X","keywords":"image classification;image recognition;ubiquitous computing;3-D acceleration signals;MFCC;compact cascaded classifier;dynamic look-up table;image processing;multiaxial Haar-like feature classifier;pervasive computing;redundant feature selection;speech classification;ubiquitous computing;versatile recognition algorithm;Acceleration;Costs;Data mining;Dynamic programming;Frequency;Humans;Image recognition;Signal processing;Speech;Table lookup;Cascaded classifier;Haar-like feature;versatile recognition","timestamp":"2014.12.22","bibtex":"@Article{Nishimura2010,\n Title = {Multiaxial Haar-Like Feature and Compact Cascaded Classifier for Versatile Recognition},\n Author = {Nishimura, J. and Kuroda, T.},\n Journal = {Sensors Journal, IEEE},\n Year = {2010},\n\n Month = {Nov},\n Number = {11},\n Pages = {1786-1795},\n Volume = {10},\n\n Abstract = {A versatile recognition algorithm has been proposed to process image, sound, and 3-D acceleration signals with a common framework at low calculation cost. Firstly, a novel 1-D Haar-like feature is used to roughly extract frequency information from temporal signals. Biaxial and mean-embedded Haar-like features are proposed to extract the standard deviation and the interaxial correlation from 3-D acceleration signals. Secondly, two techniques are proposed to build a compact cascaded classifier. Redundant feature selection (RFS) incorporates the features which are already selected in previous stage classifiers to reduce the calculation cost. A dynamic look-up table (DLUT) is proposed to construct a look-up table-based weak classifier with the smallest possible number of bins. A train loss function is by globally optimized using dynamic programming. The proposed algorithm is tested experimentally on speech/nonspeech classification and human activity recognition. The proposed algorithm yields a speech/nonspeech classification performance comparable to the state-of-art method called MFCC while reducing the calculation cost by 100 times. The algorithm also achieves human activity recognition accuracy of 96.1% with calculation cost reduction of 84% compared with the state-of-art method based on C4.5 decision-tree classifier using the basic statistical features. The proposed algorithm has been employed to build the versatile recognition processor.},\n Doi = {10.1109/JSEN.2010.2049740},\n ISSN = {1530-437X},\n Keywords = {image classification;image recognition;ubiquitous computing;3-D acceleration signals;MFCC;compact cascaded classifier;dynamic look-up table;image processing;multiaxial Haar-like feature classifier;pervasive computing;redundant feature selection;speech classification;ubiquitous computing;versatile recognition algorithm;Acceleration;Costs;Data mining;Dynamic programming;Frequency;Humans;Image recognition;Signal processing;Speech;Table lookup;Cascaded classifier;Haar-like feature;versatile recognition},\n Timestamp = {2014.12.22}\n}\n\n","author_short":["Nishimura, J.","Kuroda, T."],"key":"Nishimura2010","id":"Nishimura2010","bibbaseid":"nishimura-kuroda-multiaxialhaarlikefeatureandcompactcascadedclassifierforversatilerecognition-2010","role":"author","urls":{},"keyword":["image classification;image recognition;ubiquitous computing;3-D acceleration signals;MFCC;compact cascaded classifier;dynamic look-up table;image processing;multiaxial Haar-like feature classifier;pervasive computing;redundant feature selection;speech classification;ubiquitous computing;versatile recognition algorithm;Acceleration;Costs;Data mining;Dynamic programming;Frequency;Humans;Image recognition;Signal processing;Speech;Table lookup;Cascaded classifier;Haar-like feature;versatile recognition"],"downloads":0},"search_terms":["multiaxial","haar","feature","compact","cascaded","classifier","versatile","recognition","nishimura","kuroda"],"keywords":["image classification;image recognition;ubiquitous computing;3-d acceleration signals;mfcc;compact cascaded classifier;dynamic look-up table;image processing;multiaxial haar-like feature classifier;pervasive computing;redundant feature selection;speech classification;ubiquitous computing;versatile recognition algorithm;acceleration;costs;data mining;dynamic programming;frequency;humans;image recognition;signal processing;speech;table lookup;cascaded classifier;haar-like feature;versatile recognition"],"authorIDs":[],"dataSources":["iCsmKnycRmHPxmhBd"]}