基于神经网络的语音识别在wince上的实现和优化-电路与系统专业论文.docxVIP

优秀毕业论文 精品参考文献资料 上海 上海交通大学硕士学位论文 A bstract Language is one ofthe m ostim portantw ays to exchange inform ation am ong the m ankind.W ith the continuous developm entof com puter technology and im provem entof artificialintelligent,people expectto m ake the com puter recognize hum an language.This requirem entm akes the technique ofspeech recognition have im m ense space to develop.U p to now,m ostspeech recognition is based on conventionallinearsystem theory,such as Dynam ic Tim e W arping (D TW ) and Hidden M arkov M odel(H M M ).W ith the deep study ofspeech recognition,it is found that speech signalis a com plex nonlinearprocess.To break through existing study,nonlinearsystem theory m ethod m ustbe introduced to it.W ith the developm entofartificialneuralnetw orks (A N N ),it is possible to apply these nonlinear-system theories to speech recognition. This paperstudied speech recognition based on A NN ,com puting validation,Perform ance analysis and results assessing are handled to each partofspeech recognition process.A softw are platform w as setup based on B P neuralnetw ork m odeland a new w ay ofm ixing param eters of M FC C and L PC C was proposed.A n application ofspeech verification w as developed and optim ized on W inC E which can recognize0 to 9 ata rate of88.4% . Keywords: Speech recognition,N euralnetw ork,W inC E,eBox 上海 上海交通大学硕士学位论文 目录 第一章 绪论.............................................................................................................................................................1 1.1 课题背景 .................................................................................................................................................1 1.1.1 语音识别概述 ...............................................................................................................................1 1.1.2 数字语音识别研究的意义.........................................................................................................2 1.2 语音识别的发展和现状..........