本文選題：腦電　切入點(diǎn)：失匹配負(fù)波　出處：《清華大學(xué)》2013年博士論文　論文類型：學(xué)位論文

【摘要】：聽力障礙為患者的日常生活帶來(lái)了極大困擾，評(píng)估聽覺認(rèn)知能力可為臨床疾病診斷和認(rèn)知科學(xué)研究提供重要依據(jù)。本文基于腦電分析方法研究了聽覺認(rèn)知相關(guān)的若干問題。本文提出了從較少試次腦電數(shù)據(jù)中提取失匹配負(fù)波（Mismatchnegativity, MMN）的算法。利用事件相關(guān)電位（Event-related potential, ERP）和Gamma節(jié)律研究了噪聲條件下視覺信息對(duì)語(yǔ)音識(shí)別的影響，并采用時(shí)空域結(jié)合分析的方法研究了語(yǔ)音識(shí)別時(shí)視聽覺多感官增益成分的時(shí)間響應(yīng)和空間定位。最后，初步探討了臨床中風(fēng)患者噪聲條件下的語(yǔ)音識(shí)別能力。 提出了結(jié)合重采樣差分和最大化信噪比的rdSIM算法，可以從較少試次腦電數(shù)據(jù)中提取MMN波形。算法能夠最大化ERP信噪比，同時(shí)利用非參數(shù)重采樣降低樣本方差。差分可有效增強(qiáng)偏差和標(biāo)準(zhǔn)刺激響應(yīng)的差異成分，直接提取差異波。結(jié)果表明，rdSIM算法能夠從50個(gè)試次的偏差刺激數(shù)據(jù)中提取MMN波形，所得波形符合MMN的典型性質(zhì)。 基于腦電ERP研究不同信噪比環(huán)境下視覺輸入對(duì)語(yǔ)音識(shí)別的影響。在-16、-12、-8、-4和0dB五種信噪比條件下，發(fā)現(xiàn)視聽覺多感官增益隨信噪比變化呈現(xiàn)倒U型，最大增益在信噪比-12dB。證實(shí)不同信噪比條件下視聽覺多感官增益不完全符合反轉(zhuǎn)原則，而與跨模態(tài)的隨機(jī)共振現(xiàn)象一致。 基于腦電Gamma節(jié)律信號(hào)研究不同信噪比環(huán)境下視聽覺多感官增益的變化規(guī)律。發(fā)現(xiàn)Gamma頻段視聽覺相比單聽的多感官增益在-12dB最大。在40~100ms時(shí)間窗內(nèi)，視聽覺Gamma節(jié)律的相位鎖定能量、總能量和相位鎖定因子同步增強(qiáng)，這表明Gamma節(jié)律的相位同步發(fā)揮著整合不同感官信息的作用。 提出了結(jié)合rdSIM時(shí)空域模式和等效偶極子源分析的時(shí)空分析方法。采用rdSIM算法提取多感官增益的波形，，證實(shí)變化較大的幅度出現(xiàn)在N1/P2成分。再利用源定位技術(shù)分析多感官增益的腦區(qū)分布，結(jié)果證實(shí)多感官增益的腦活動(dòng)分布在緣上回、角回、內(nèi)頂葉溝和顳上溝。 基于腦電ERP研究聲音噪聲環(huán)境下中風(fēng)患者的語(yǔ)音識(shí)別能力。行為學(xué)和ERP的初步分析結(jié)果表明失語(yǔ)癥患者視聽覺多感官增益大于偏癱非失語(yǔ)患者。提示失語(yǔ)患者可以從多感官增益中獲益更大。
[Abstract]:Hearing impairment brings great trouble to patients' daily life. The evaluation of auditory cognitive ability can provide important basis for clinical disease diagnosis and cognitive science research. This paper studies some problems related to auditory cognition based on EEG analysis. An algorithm for extracting mismatched negative waves (MMN) is proposed. The effects of visual information on speech recognition under noise are studied by using event-related potentialand Gamma rhythms. The temporal response and spatial localization of the multi-sensory gain component of audiovisual perception in speech recognition were studied by the method of spatio-temporal domain combined with analysis. Finally, the speech recognition ability of clinical apoplexy patients under noise condition was preliminarily discussed. A rdSIM algorithm combining resampling difference and maximum SNR is proposed, which can extract MMN waveform from less subEEG data. The algorithm can maximize the SNR of ERP. At the same time, the nonparametric resampling is used to reduce the sample variance. Difference can effectively enhance the difference components of the deviation and standard stimulus response, and extract the difference wave directly. The results show that the MMN waveform can be extracted from the bias stimulus data of 50 samples. The obtained waveforms accord with the typical properties of MMN. The effects of visual input on speech recognition in different SNR environments were studied based on EEG ERP. Under five signal-to-noise ratio (SNR) conditions, it was found that the multi-sensory gain of audiovisual sense was inversely U-shaped under five signal-to-noise ratios (SNR) of -16 ~ (-12) -12 ~ (-8) ~ (-4) and 0 dB. The maximum gain is in the SNR of -12 dB. It is proved that the multi-sensory gain of the audiovisual sense does not fully conform to the inversion principle under different SNR conditions, but is consistent with the cross-modal stochastic resonance phenomenon. Based on EEG Gamma rhythmic signals, the variation of multisensory gain of audiovisual sense in different SNR environments was studied. It was found that the multisensory gain in Gamma band was maximum at -12 dB compared with that in single hearing, and in the time window of 40 ~ 100ms. The phase locking energy, total energy and phase locking factor of Gamma rhythm of audiovisual sense are enhanced synchronously, which indicates that phase synchronization of Gamma rhythm plays a role in integrating different sensory information. A spatio-temporal analysis method based on rdSIM spatio-temporal model and equivalent dipole source analysis is proposed. The multi-sensory gain waveform is extracted by rdSIM algorithm. The results showed that the distribution of multi-sensory gain was found in the superior marginal gyrus, the angular gyrus, the internal parietal sulcus and the superior temporal sulcus by using the source location technique, and the results showed that the brain activity of the multi-sensory gain was distributed in the superior marginal gyrus, the angular gyrus, the internal parietal sulcus and the superior temporal sulcus. The results of behavioral and ERP analysis showed that the auditory and visual multisensory gain of aphasia patients was higher than that of hemiplegic non-aphasia patients. The results suggested that aphasia patients had higher multi-sensory gain than those with hemiplegic non-aphasia. You can benefit more from multiple sensory gains.
【學(xué)位授予單位】：清華大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2013
【分類號(hào)】：R764;TN911.7

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 畢華;梁洪力;王玨;;重采樣方法與機(jī)器學(xué)習(xí)[J];計(jì)算機(jī)學(xué)報(bào);2009年05期

2 王靜;李小俚;邢國(guó)剛;萬(wàn)有;;Gamma神經(jīng)振蕩產(chǎn)生機(jī)制及其功能研究進(jìn)展[J];生物化學(xué)與生物物理進(jìn)展;2011年08期

本文編號(hào)：1639429