Residual Attention Based Long Short-Term Memory Based Text Attributes of Intangible Cultural Heritage

The significant National cultural treasure is Intangible Cultural Heritage (ICH) which has some issues in sustainability and inheritance. With the growth of digital technique, maximizing application and research of digital technique in ICH is represented. In this research, the standardized and scientific audio data dedicated for Cantonese Open is employed and classification technique for Cantonese open singing genres dependent on Residual Attention based Long Short-Term memory (Residual based LSTM) method is proposed. Proposed method provides resemblance of rhythm features of various Cantonese opera singing genres. Actual signal is pre-processed for obtaining a Mel-Frequency Cepstrum as method's input. Cascade combination integrated every shallow segment and deep features. The residual based LSTM method is a hybridization network that improves contextual relevance among signals. The proposed method attained intelligence classification of Cantonese opera information, while efficiently resolve the issue which previous techniques are challenging for classification. The proposed method attained a precision of 93.61%, recall of 92.74% and F1-measure of 92.95% which is more efficient than other existing methods.