Abstract:
PURPOSE: A distance function combination method for an information detection system is provided to improve the performance of a matching engine by applying a matching system to a QbSH(Query-by-Singing or Humming) system. CONSTITUTION: An N length signal vector including a q1 or a qn element value and an M length signal vector including a p1 or a pm element value are inputted(S500). A distance value between element values is calculated by using a distance function(S502). An MXN matrix is calculated by using the calculated distance value(S506). A distance between the two signal vectors is calculated by using information including a value which constructs a first line of the MXN matrix. A value which constructs an M line of the MXN matrix is a distance value between an element and a vector. [Reference numerals] (S500) Collecting training data; (S502) Defining a distance and f(·); (S504) Determining the f(·); (S506) Determining d(·); (S508) Using in DTW
Abstract:
PURPOSE: A humming based audio searching system and method thereof are provided to enable a user to search desired sound by comparing user humming with sound stored in a data base. CONSTITUTION: A feature extraction apparatus(120) extracts feature information by inputting humming. A matching apparatus(130) respectively calculates matching values between the humming and a sound source stored in a database by using a pitch based LS(Linear Scaling), a pitch base DTW(Dynamic Time Warping), and a QB(Quantized Binary) code based LS method. A combination apparatus(140) calculates integrated values by applying multiplication to the matching values. A searching apparatus(150) searches the sound source similar with the humming by using the integrated values.
Abstract:
PURPOSE: A preference information management system for web-based multimedia content is provided to automatically extracts user history information and user preference according to the individual content based on one basic targeting service algorithm. CONSTITUTION: A multi-contents metadata module(100) stores metadata about plural contents genres. A user preference update module(110) collects the usage pattern of a user. The user preference update module classifies preference patterns or non-preference patterns. The user preference update module updates user preference information based on preference patterns or non-preference patterns.