【摘要】：在過去的2016年,人工智能、虛擬現(xiàn)實(shí)、可穿戴設(shè)備等已成為科技行業(yè)研究的前沿和熱點(diǎn),這些研究都不可避免的需要人與計(jì)算機(jī)進(jìn)行交互,語(yǔ)音比鍵盤鼠標(biāo)的交互方式有更高的效率,且語(yǔ)音有復(fù)雜的情感表達(dá),對(duì)交互的體驗(yàn)有很大的提升。因此語(yǔ)音識(shí)別技術(shù)必將作為人機(jī)交互最便捷的方式而被廣泛應(yīng)用。長(zhǎng)期以來,在語(yǔ)音識(shí)別領(lǐng)域聲學(xué)模型的建模都是使用GMM-HMM模型,該模型具有可靠的精度,并且有成熟的EM算法來進(jìn)行模型參數(shù)訓(xùn)練,因此GMM-HMM模型廣泛應(yīng)用在語(yǔ)音識(shí)別領(lǐng)域。但因?yàn)镚MM模型屬于淺層模型,隨著數(shù)據(jù)量的增加建模能力明顯不足。深度神經(jīng)網(wǎng)絡(luò)(DNN)因其對(duì)復(fù)雜數(shù)據(jù)有更好的建模與學(xué)習(xí)能力,成為語(yǔ)音識(shí)別領(lǐng)域研究的熱點(diǎn)。本文深入研究了基于HMM模型和DNN模型的識(shí)別算法,分析兩個(gè)模型的優(yōu)點(diǎn)以及不足,主要進(jìn)行了以下工作:(1)對(duì)基于隱馬爾科夫模型(HMM)的語(yǔ)音識(shí)別算法進(jìn)行深入研究,并使用CMUSphinx語(yǔ)音識(shí)別平臺(tái)構(gòu)建一個(gè)機(jī)器人控制命令語(yǔ)音識(shí)別系統(tǒng),對(duì)機(jī)器人十個(gè)控制命令的語(yǔ)音信號(hào)進(jìn)行訓(xùn)練得到語(yǔ)言模型和聲學(xué)模型。實(shí)驗(yàn)解碼結(jié)果表明,該系統(tǒng)平均錯(cuò)詞率為7.1%,具有良好的識(shí)別效果,在小詞匯量漢語(yǔ)語(yǔ)音識(shí)別中具有較高的識(shí)別率。(2)針對(duì)HMM模型的不足,對(duì)深度神經(jīng)網(wǎng)絡(luò)中的深度信念網(wǎng)絡(luò)(DBN)深入研究,使用Kaldi語(yǔ)音識(shí)別工具實(shí)現(xiàn)了大詞匯量中文連續(xù)語(yǔ)音識(shí)別系統(tǒng)的構(gòu)建,對(duì)中文開源語(yǔ)音庫(kù)THCHS30進(jìn)行DNN聲學(xué)模型訓(xùn)練,實(shí)驗(yàn)結(jié)果表明DNN模型比三音子模型錯(cuò)詞率降低了5.79%,DNN模型在大詞匯量語(yǔ)音識(shí)別系統(tǒng)中具有更好的識(shí)別效果。同時(shí)本文使用Kaldi對(duì)TIMIT語(yǔ)音庫(kù)訓(xùn)練得到大詞匯量英文語(yǔ)音識(shí)別系統(tǒng),取得了較高的識(shí)別率。(3)噪聲干擾一直是語(yǔ)音識(shí)別的難點(diǎn),在使用Kaldi進(jìn)行聲學(xué)模型訓(xùn)練的過程中,通過在訓(xùn)練和測(cè)試語(yǔ)音加入白噪聲、汽車背景噪聲、自助餐背景噪聲進(jìn)行DNN訓(xùn)練,并與多種模型對(duì)比,實(shí)驗(yàn)結(jié)果表明DAE模型在低維表示方面具有更好的效果,可以用于恢復(fù)噪聲損壞的輸入。
[Abstract]:In the past 2016, artificial intelligence, virtual reality, wearable devices and so on have become the frontier and hot spot of the technology industry research, these research inevitably need people and computer interaction, Speech is more efficient than keyboard and mouse, and speech has complex emotion expression, so the interaction experience is greatly improved. Therefore, speech recognition technology will be widely used as the most convenient way of human-computer interaction. For a long time, the modeling of acoustic models in the field of speech recognition is based on GMM-HMM model, which has reliable precision and mature EM algorithm to train the model parameters. Therefore, GMM-HMM model is widely used in the field of speech recognition. However, because GMM model belongs to shallow model, the ability of modeling is obviously insufficient with the increase of data volume. Deep neural network (DNN) has become a hot topic in speech recognition field because of its better modeling and learning ability for complex data. In this paper, the recognition algorithms based on HMM model and DNN model are deeply studied, and the advantages and disadvantages of the two models are analyzed. The main work is as follows: (1) the speech recognition algorithm based on Hidden Markov Model (HMM) is studied deeply. A robot control command speech recognition system is constructed by using CMUSphinx speech recognition platform, and the speech model and acoustic model are obtained by training the speech signal of the robot's ten control commands. The experimental results show that the average error rate of the system is 7.1, which has a good recognition effect, and has a high recognition rate in small vocabulary Chinese speech recognition. (2) aiming at the deficiency of HMM model, In this paper, the deep belief network (DBN) in depth neural network is deeply studied, the large vocabulary Chinese continuous speech recognition system is constructed by using Kaldi speech recognition tool, and the DNN acoustic model training is carried out on THCHS30, a Chinese open source speech database. The experimental results show that the DNN model has a better recognition effect in the large vocabulary speech recognition system than the trisyllabic model. At the same time, this paper uses Kaldi to train the TIMIT speech corpus to obtain a large vocabulary English speech recognition system, and obtains a high recognition rate. (3) noise interference is always a difficult point in speech recognition. In the process of using Kaldi to train acoustic model, By adding white noise, automobile background noise and buffet background noise into the training and testing speech, the DNN training is carried out, and compared with many models, the experimental results show that the DAE model is more effective in low dimensional representation. Can be used to restore noise damaged input.
【學(xué)位授予單位】：江西理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TN912.34

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 劉旺玉;SHIRAISHI HIROSHI;;基于GMM-HMM和深層循環(huán)神經(jīng)網(wǎng)絡(luò)的復(fù)雜噪聲環(huán)境下的語(yǔ)音識(shí)別[J];制造業(yè)自動(dòng)化;2016年05期

2 屈丹;張文林;;基于本征音子說話人子空間的說話人自適應(yīng)算法[J];電子與信息學(xué)報(bào);2015年06期

3 王山海;景新幸;楊海燕;;基于深度學(xué)習(xí)神經(jīng)網(wǎng)絡(luò)的孤立詞語(yǔ)音識(shí)別的研究[J];計(jì)算機(jī)應(yīng)用研究;2015年08期

4 尹寶才;王文通;王立春;;深度學(xué)習(xí)研究綜述[J];北京工業(yè)大學(xué)學(xué)報(bào);2015年01期

5 戴禮榮;張仕良;;深度語(yǔ)音信號(hào)與信息處理:研究進(jìn)展與展望[J];數(shù)據(jù)采集與處理;2014年02期

6 余凱;賈磊;陳雨強(qiáng);徐偉;;深度學(xué)習(xí)的昨天、今天和明天[J];計(jì)算機(jī)研究與發(fā)展;2013年09期

7 陸俊;張瓊;楊俊安;王一;劉輝;;嵌入深度信念網(wǎng)絡(luò)的點(diǎn)過程模型用于關(guān)鍵詞檢出[J];信號(hào)處理;2013年07期

8 謝怡寧;黃金杰;何勇軍;;噪聲環(huán)境下智能機(jī)器人語(yǔ)音控制特征提取方法[J];北京郵電大學(xué)學(xué)報(bào);2013年03期

9 楊雅婷;馬博;王磊;吐爾洪·吾司曼;李曉;;維吾爾語(yǔ)語(yǔ)音識(shí)別中發(fā)音變異現(xiàn)象[J];清華大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年09期

10 孫峰;姚毅;李成剛;;LM算法在神經(jīng)網(wǎng)絡(luò)語(yǔ)音識(shí)別中的應(yīng)用[J];科學(xué)技術(shù)與工程;2011年09期

相關(guān)碩士學(xué)位論文 前3條

1 王琳;噪聲環(huán)境下的魯棒語(yǔ)音識(shí)別技術(shù)研究[D];哈爾濱工業(yè)大學(xué);2016年

2 張建華;基于深度學(xué)習(xí)的語(yǔ)音識(shí)別應(yīng)用研究[D];北京郵電大學(xué);2015年

3 陳碩;深度學(xué)習(xí)神經(jīng)網(wǎng)絡(luò)在語(yǔ)音識(shí)別中的應(yīng)用研究[D];華南理工大學(xué);2013年

，

本文編號(hào)：2324964