改进黑翅鸢算法的LSTM永嘉方言识别策略

打开文本图片集

中图分类号：TP391 文献标识码：A 文章编号：1006-8228（2025）10-44-07

Abstract：AimingattheaentandtonalcharacteristicsoftheYongjiadialect，thispaperproposesaLongShort-TemMemory （LSTM）speechrecognitionmodelbasedonanImprovedBlack-wingedKiteOptimizationAlgorithm（BKA）.Firstlytoddreste shortcomingsofthetraditionalBlack-wingedKiteOptimizationAgorithm（BKA），theimprovedalgorithmintroducestheCebshev chaoticsequecetooptiizeitspopulationiitializationprocessDuringteositionupdatephaseitincoporatesalsositio basedlearningstrategytobalancethealgorithm'sglobalexplorationandlocalexploitationcapabilities，therebyenhancingits convergenceacuracyandspeed.Topreventprematureconvergence，animprovedGaussanmutationfactorisaddedtointroduce apropriateperturbations，helpingthealgorithmescapefromlocaloptima.Secondly，theimprovedBlack-wingedKiteAlgorithis integratedwiththeLSTMnetworktoconstructanIBKA-LSTMspeechrecognitionmodelFinallMel-FrequencyCepstral Coeficients（MFCC）areusedtoextractfeaturesfromthecharacteristcdialectspeechandrecognitionisperformedbytheIBKALSTM model.

Keywords：Black-wingedKiteAlgorithm；ChebyshevChaoticSequence;LensOpposition-BasedLearning；SpechRecognition; LSTM Network

0引言

伴随人工智能技术的发展以及语音识别技术的快速进步，众多语言学者致力于建立完整的大规模方言标准库，对各地的方言进行识别和分析。（剩余9032字）