《電子技術(shù)應(yīng)用》
您所在的位置:首頁(yè) > 其他 > 設(shè)計(jì)應(yīng)用 > 融合CNN和Transformer編碼器的變聲語(yǔ)音鑒別與還原
融合CNN和Transformer編碼器的變聲語(yǔ)音鑒別與還原
信息技術(shù)與網(wǎng)絡(luò)安全 1期
魏春雨,,孫 蒙,,劉 偉,,張星昱
(陸軍工程大學(xué) 指揮控制工程學(xué)院,,江蘇 南京210007)
摘要: 語(yǔ)音變聲偽裝會(huì)導(dǎo)致人耳感知和聲紋識(shí)別出現(xiàn)錯(cuò)誤,,從而達(dá)到隱匿說(shuō)話人真實(shí)身份的目的。為削弱變聲語(yǔ)音的影響,,提出一種融合卷積神經(jīng)網(wǎng)絡(luò)(Convolutional Neural Networks,,CNN)和Transformer編碼器的模型,提取變聲語(yǔ)音的局部特征和全局特征用于判別變聲因子,,并根據(jù)變聲因子的數(shù)值實(shí)施變聲語(yǔ)音還原,。在中英文真實(shí)場(chǎng)景錄音數(shù)據(jù)集上驗(yàn)證了所提方法的有效性,對(duì)變聲因子判別實(shí)現(xiàn)了95%以上的準(zhǔn)確率,。利用所提出的方法,,在黑箱條件下對(duì)某型商用硬件變聲器輸出的語(yǔ)音進(jìn)行鑒別與還原,取得了較好的效果,。
中圖分類號(hào): TP391.9
文獻(xiàn)標(biāo)識(shí)碼: A
DOI: 10.19358/j.issn.2096-5133.2022.01.008
引用格式: 魏春雨,,孫蒙,劉偉,,等. 融合CNN和Transformer編碼器的變聲語(yǔ)音鑒別與還原[J].信息技術(shù)與網(wǎng)絡(luò)安全,,2022,41(1):47-54.
Identification and restoration of transformed voice by fusing CNN and Transformer encoder
Wei Chunyu,,Sun Meng,,Liu Wei,Zhang Xingyu
(College of Command and Control Engineering,,Army Engineering University of PLA,,Nanjing 210007,China)
Abstract: Voice transformation will lead to errors in auditory perception and speaker recognition so as to conceal the speaker′s real identity. In order to reduce the negative impact of transformed voice, a model fusing Convolutional Neural Networks(CNN) and Transformer encoder was proposed in this paper, which extracted local and global features of transformed voice to predict the disguise factor, and restored the original voice according to the value of the disguise factor. The validity of the proposed method was verified on datasets of both Chinese and English recorded in real-world scenes, where the accuracy was higher than 95%. Under the condition of black box, the proposed method had good performance when identifying and restoring the output voice of a commercial hardware of voice changer.
Key words : pitch scaling,;voice anti-disguise,;voice restoration;time-frequency features

0 引言

近年來(lái),,基于指紋,、人臉、虹膜,、聲紋的生物特征識(shí)別技術(shù)得到了迅速發(fā)展[1],。其中,聲紋識(shí)別[2]以其采集方便,、無(wú)需直接接觸,、可體現(xiàn)說(shuō)話人的主觀意圖等優(yōu)點(diǎn),,被廣泛應(yīng)用于身份認(rèn)證,、刑事偵查等領(lǐng)域。公安部門可以通過(guò)提取犯罪嫌疑人的聲紋信息,,將其與數(shù)據(jù)庫(kù)中的聲紋信息進(jìn)行比對(duì),,進(jìn)而確認(rèn)嫌疑人的真實(shí)身份。然而,這種確認(rèn)經(jīng)常受到各種因素的干擾,,其中語(yǔ)音變聲偽裝就可以通過(guò)隱匿說(shuō)話人的身份,,降低聲紋確認(rèn)的效果。

與通過(guò)壓低嗓音,、捏鼻子等方式物理干擾人的發(fā)聲器官進(jìn)行的語(yǔ)音偽裝不同,,語(yǔ)音變聲偽裝是利用變聲工具對(duì)語(yǔ)音參數(shù)進(jìn)行修改實(shí)現(xiàn)變聲[3]。目前,,各種硬件變聲設(shè)備和變聲軟件在匿名采訪,、網(wǎng)絡(luò)聊天、電子游戲等場(chǎng)景中得到了廣泛應(yīng)用,。同時(shí),,一些不法分子利用變聲語(yǔ)音進(jìn)行違法犯罪活動(dòng),對(duì)案件的偵破造成了干擾,。相比通過(guò)物理干擾發(fā)聲器官而產(chǎn)生的偽裝語(yǔ)音,,變聲語(yǔ)音更加難以辨認(rèn)。變聲工具會(huì)改變語(yǔ)音的頻譜結(jié)構(gòu),,進(jìn)而影響聲紋識(shí)別模型對(duì)說(shuō)話人的驗(yàn)證[4-5],。有研究表明,向普通人播放經(jīng)過(guò)變聲偽裝的語(yǔ)音,,甚至不能準(zhǔn)確地判斷出說(shuō)話人的性別,。



本文詳細(xì)內(nèi)容請(qǐng)下載:http://forexkbc.com/resource/share/2000003936






作者信息:

魏春雨,,孫  蒙,,劉  偉,張星昱

(陸軍工程大學(xué) 指揮控制工程學(xué)院,,江蘇 南京210007)


此內(nèi)容為AET網(wǎng)站原創(chuàng),,未經(jīng)授權(quán)禁止轉(zhuǎn)載。