【摘要】：針對(duì)包含環(huán)境噪聲和信道失真等噪聲的語音處理問題,提出了一種基于自適應(yīng)心理聲學(xué)模型的智能語音識(shí)別系統(tǒng),并建立了聽覺模型.該模型將心理聲學(xué)和耳聲發(fā)射(OAE)合并到了自動(dòng)語音識(shí)別(ASR)系統(tǒng)中,利用AURORA2數(shù)據(jù)庫分別在清潔訓(xùn)練條件和多訓(xùn)練條件下進(jìn)行試驗(yàn).結(jié)果表明,所提出的特征提取方法可以顯著提高詞識(shí)別率,優(yōu)于梅爾頻率倒譜系數(shù)(MFCC)、前向掩蔽(FM)、側(cè)向抑制(LI)和倒譜平均值及方差歸一化(CMVN)算法,能夠有效地提高智能語音識(shí)別系統(tǒng)的性能.
[Abstract]:In this paper, an intelligent speech recognition system based on adaptive psychoacoustic model is proposed to deal with speech processing problems including ambient noise and channel distortion, and an auditory model is established. The model combines psychoacoustics and otoacoustic emission (OAE) (OAE) into automatic speech recognition (ASR) system. The AURORA2 database is used to test under the condition of clean training and multiple training respectively. The results show that the proposed feature extraction method can significantly improve the word recognition rate, and is better than the Mel frequency cepstrum coefficient (MFCC), forward masking (FM), lateral suppression (LI), cepstrum mean and variance normalized (CMVN) algorithm. It can effectively improve the performance of the intelligent speech recognition system.
【作者單位】： 南昌大學(xué)科學(xué)技術(shù)學(xué)院;
【基金】：江西省教育廳科學(xué)技術(shù)研究項(xiàng)目(GJJ151504,GJJ151505) 江西省教育改革課題資助項(xiàng)目(JXJG-14-28-3,JXJG-14-28-1,JXJG-14-28-6,JXJG-14-28-8)
【分類號(hào)】：TN912.34

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 汪洪波;;高效的語音識(shí)別系統(tǒng)[J];中國物流與采購;2006年12期

2 汪洪波;;語音識(shí)別系統(tǒng)在配送中心的應(yīng)用[J];信息與電腦;2006年06期

3 楊q，

本文編號(hào)：2250440