TWI270605B - Voiceprint key system - Google Patents
Voiceprint key system Download PDFInfo
- Publication number
- TWI270605B TWI270605B TW94119966A TW94119966A TWI270605B TW I270605 B TWI270605 B TW I270605B TW 94119966 A TW94119966 A TW 94119966A TW 94119966 A TW94119966 A TW 94119966A TW I270605 B TWI270605 B TW I270605B
- Authority
- TW
- Taiwan
- Prior art keywords
- voiceprint
- voice
- code
- key
- feature
- Prior art date
Links
Landscapes
- Storage Device Security (AREA)
- Lock And Its Accessories (AREA)
Abstract
Description
1270605 94. 08· 08第94119966號專利說明書及圖式修正本 九、發明說明: 【發明所屬之技術領域】 本發明係關於-種聲紋密碼鎖系統,其特別有關於 於電子資料〔職或資料夹〕之聲贿碼齡統,該聲紋资 碼鎖可供固定或攜帶設定,以強化個人電子資山 【先前技術】 emt。 时彻獻生物雜分繼逐漸成為個人 辨識的主要技術。在辨識技術上,目前可供辨識的生物特ς 包,臉孔辨識、指紋辨識、掌紋辨識、聲紋辨識、眼球虹膜 辨識及去氧核糖核酸〔DNA〕排序配對辨識等。 、 -般針對電子㈣安全保護祇個人密碼加以 義人密碼無法提供有朗安全賴,例如密碼外鴻、 着=解4。因此’電子資料安全賴必鮮求其他保護措施, 藉由強化個人身分辨識之功能,以提升保護電子資料安全。 右基於考量辨識技術的架構成本及實際操作時,可發 辨識相對較適合選擇做為個人身分辨識之用途。Χ 、,、 ㈣聲紋驗證方法,如巾軌國專利1270605 94. 08· 08 No. 94119966 Patent Specification and Drawing Revisions IX. EMBODIMENT OF THE INVENTION: TECHNICAL FIELD OF THE INVENTION The present invention relates to a voiceprint lock system, which is particularly relevant to electronic data. The data folder] is a code of bribe code, and the sound code code lock can be fixed or carried to set up to strengthen the personal electronic Zishan [previous technology] emt. The time to provide biological diversity has gradually become the main technology for personal identification. In terms of identification technology, currently available biometrics, face recognition, fingerprint recognition, palmprint recognition, voiceprint recognition, eye iris recognition, and DNA sequencing matching identification. - General for electronic (four) security protection only personal passwords. Human rights passwords can not provide security, such as passwords, and = solution 4. Therefore, electronic data security will require other protection measures to enhance the protection of electronic data by strengthening the function of personal identification. Right based on the cost of the architecture and the actual operation of the identification technology, it can be identified as a more suitable choice for personal identity identification. Χ,,, (4) Voiceprint verification methods, such as the patent
利用聲譜資訊辯識使用者的方法與其裝置」發明專:5, J 使者特有的聲譜資訊辨識使用者的身份,以決定 使用者疋碰職權。财法包含步^⑴ 語=;〔2〕、自該語音的聲譜中取= 徵作為-參考樣本、’若「是」則將該語音特 ㈣本冋時奴—界限,S「 y驟;⑷、將該語音特徵與參 ^ 、 依該峨果物者之·轉;〔6〕=計=果 C:\LiiKto\nt Patvpugsoi-iooooxpicMia.doc 1270605 94· 08· 08第94119966號專利說明書及圖式修正本 與設定界限比較;〔7〕、依該比較結果決定該使用者是否為 一授權使用者。該方法係使用於行動電話,其利用聲譜分析 方法將語音之獨特資訊取出,藉此進行辨識使用者之方法。 遠第490655號主要利用每一時框〔frame〕之主要值與使用 者設定的界限進行比較,蚊語音之始點與終點後,再利用 Princen-Bradley濾波器轉換已偵測的語音訊號,以便取得 其對應聲譜圖案。該聲譜圖案翻先儲存之參考聲譜樣本進 行比對,以辨識使用者之聲紋。The use of sound spectrum information to identify the user's method and its device" Invention: 5, J The unique spectral information of the messenger identifies the user's identity to determine the user's ability to bump. The financial method includes the step ^(1) ==[2], from the sound spectrum of the speech, the sign is taken as the reference sample, and if it is "yes", the voice is specially (four) the time slave - the limit, S " y (4), the voice features and parameters, according to the fruit of the person; [6] = count = fruit C: \ LiiKto \ nt Patvpugsoi-iooooxpicMia.doc 1270605 94 · 08 · 08 No. 94119966 patent specification and The schema revision is compared with the set limit; [7], according to the comparison result, whether the user is an authorized user. The method is used in a mobile phone, which uses sound spectrum analysis method to take out the unique information of the voice, borrowing This method is used to identify the user. Far 490655 mainly uses the main value of each frame to compare with the limit set by the user. After the start and end points of the mosquito voice, the Princen-Bradley filter is used to convert The detected voice signal is obtained to obtain a corresponding sound spectrum pattern, and the sound spectrum pattern is compared with the reference sound spectrum sample stored first to identify the voiceprint of the user.
簡言之’該第490655號需要進行圖案的匹配及距離的 運算,若該運算距離未超過界限時,使用者即可通過聲紋辨 識。然而:該第490655號在進行圖案的匹配及距離的運算時, 必須計算在參考樣本及職樣本之間的距離。事實上,該失 =樣本所侧資料庫的空_#大,因此其不但需要較大的 貝料庫如且.需要更長的檔轉輸時間。在個人電子 ==_^_蛛刪爾在儲存空 之佔考樣本 一 辟參考樣本之資料庫空間, 以便適麟承触__有_猶顧或裝置。 上述之缺點,其制—聲紋驗證 糸、、先,在擷取—特徵之前,彻 另_料處理量及提升辨識率。 另外柄明利用該聲紋驗證系 〔voiceprintkey〕,例如選取-一 C:\Linda\PK Pat\PKWfl卜丨咖0\ΡΚ9β丨3. doc 1270605 94· 08. 08第94119966號專利說明書及囷式修正本 鑰地將一電子資料進行加密〔encrypt〕或解密〔decrypt〕 運算’以構成一聲紋密碼鎖,該聲紋密碼鎖用以保護該電子 資料。 【發明内容】 本發明主要目的係提供一種聲紋密碼鎖系統,其利用一 聲紋鑰匙將一電子資料進行加密或解密運算,以構成一聲紋 密碼鎖,使本發明具有提升電子資料安全之功效。 ^ 本發明次要目的係提供一種聲紋密碼鎖系統,其採用一 聲紋驗證系統,在訓練及測試擷取語音特徵之前,利甩前端 =理將有效§吾音資訊自原始輸人語音擷取,以濾除無效語音 資訊’使本發明具有減少資料處理量及提升辨識率之功效。 T% 本發明次要目的係提供一種聲紋密碼鎖系統,其在測試 Ά上’ 前端處理減少有效語音資訊,並進行娜語音特徵, 並_轉比·〔 Viterbi〕演算法麟最她雜,以便計算 模型參數〔各狀態的賊值及贿數〕供儲存,在語音訓: 或測試時’料算模·數及戦語音賴之間相似機率 2 ’以獲得—聲紋特徵值,使本發明具有簡化訓練及測試 作業之功效。 ^據本發明之聲紋密碼鎖系統,該系統包含一聲紋餘 用以將―電子資料進行加密或職運算,以構成-聲 纹二聲紋驗證系統用以產生一聲紋特徵值,自該聲 聲紋錄匙。該聲紋驗證系統包含—前端處理 幹入—訓練職系統,以便對原始 輸入,σ曰-貝枓進行訓練或測試作業。 在訓練吾音上’齡丨練系統利用該前端處理部自該原始 C:\Linde\PK Ρ«\ΡΜ50Μ 0000\ΡΚ9βΙ3. doc 1270605 % Φ 08第94119966號專利說明書及囷式修正本 輸入語音資料擷取有效訓練語 進行擷取該有效訓練語音=成,再利用该特徵娜部 資訊以獲得最相似路徑,以便作該有效訓練語音 擷取有效測試語音資:=理部自該原始輸入赠 效測試語音特徵;再進 ^亥特徵操取部進行擷取該有 間相似機細便細結果』m語音概與模型參數之 【實施方式】 確被了為解了讓述和其他目的、概、和優點能更明 作詳細_如下/牛本發明較佳實施例,並配合所附圖式, 本發明較佳實施例聲紋密碼鎖系統採用一聲紋驗證系 ^以便對原始輸入語音資料進行訓練或測試作業。第i圖 揭不本發明較佳實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之流程方塊圖。 口月多…、第1圖所示,本發明較佳實施例聲紋密碼鎖系統 所_的雜驗證系統i包含—訓練系統1()及一測試系統 20,以便對原始輸入語音資料進行訓練或測試作業。該聲紋 驗證系統1另包含-前端處理部、—特徵操取部、一儲存部 及一運算部。该前端處理部及特徵擷取部供該訓練系統1〇及 測甙系統20進行前端處理及特徵擷取,該儲存部供語音特徵 加以儲存,該運算部則將該儲存語音特徵及輸入語音特徵加 以運算。 使用者進入該聲紋驗證系統1時,首先輸入帳號,以便 進行確認身分。接著,該系統依輸入帳號查詢資料庫,是否In short, the number 490655 requires pattern matching and distance calculation. If the calculation distance does not exceed the limit, the user can recognize the voiceprint. However, the No. 490655 must calculate the distance between the reference sample and the job sample when performing pattern matching and distance calculation. In fact, the loss = the empty _# of the database on the side of the sample, so it requires not only a larger library, but also a longer file transfer time. In the personal electronic ==_^_ spiders in the storage of the sample of the sample of a reference to the database space of the reference sample, so that the appropriate lining __ have _ still care or device. The above-mentioned shortcomings, the system-voiceprint verification 糸, first, before the capture-features, the amount of processing and the recognition rate. In addition, the voiceprinting system is used, for example, the selection of -C:\Linda\PK Pat\PKWfl 丨 丨 0 ΡΚ 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 270 Correcting the key to encrypt (encrypt) or decrypt (decrypt) an electronic data to form a voice code lock, the voiceprint lock is used to protect the electronic data. SUMMARY OF THE INVENTION The main object of the present invention is to provide a voiceprint password lock system that utilizes a voiceprint key to encrypt or decrypt an electronic data to form a voiceprint password lock, so that the present invention can improve the security of electronic data. efficacy. The secondary object of the present invention is to provide a voiceprint code lock system using a voiceprint verification system, which is effective before training and testing to capture voice features. §My voice information from the original input voice 撷Taking to filter out invalid speech information' enables the invention to reduce the amount of data processing and improve the recognition rate. T% The secondary objective of the present invention is to provide a voiceprint code lock system which reduces the effective voice information on the test ' front end processing, and performs the Na-voice feature, and the _ turn ratio [ Viterbi] algorithm is the most complicated. In order to calculate the model parameters [the thief value and the number of bribes in each state] for storage, in the voice training: or test, the similar probability between the calculation of the modulus and the number of voices is 2' to obtain the voiceprint feature value, so that The invention has the effect of simplifying training and testing operations. According to the voiceprint combination lock system of the present invention, the system includes a voice pattern for encrypting or electronically calculating the electronic data to form a voiceprint two-voice verification system for generating a voiceprint characteristic value. The voice recording key. The voiceprint verification system includes a front-end processing dry-training system to train or test the original input, σ曰-Bei. In the training of my voice, the 'sports training system utilizes the front-end processing unit from the original C:\Linde\PK Ρ«\ΡΜ50Μ 0000\ΡΚ9βΙ3. doc 1270605 % Φ 08 Patent No. 94191996 and the revised version of the input voice data Take effective training language to capture the effective training voice = Cheng, and then use the characteristic Na Department information to obtain the most similar path, in order to make the effective training voice capture effective test voice: = the Ministry of the original input effect Test the characteristics of the voice; then enter the feature of the feature of the feature of the similar feature, and the result of the model is the same as the other parameters. And the advantages can be more clearly described in detail below. In accordance with the preferred embodiment of the present invention, and in conjunction with the drawings, the voiceprint combination lock system of the preferred embodiment of the present invention uses a voiceprint verification system to perform the original input voice data. Training or testing assignments. Figure i is a block diagram showing the flow pattern verification system used in the voiceprint lock system of the preferred embodiment of the present invention. More than one month... As shown in Fig. 1, the miscellaneous verification system i of the voiceprint lock system of the preferred embodiment of the present invention includes a training system 1() and a test system 20 for training the original input voice data. Or test the job. The voiceprint verification system 1 further includes a front end processing unit, a feature operation unit, a storage unit, and a calculation unit. The front end processing unit and the feature extraction unit are configured to perform front end processing and feature extraction on the training system 1 and the measurement system 20, wherein the storage unit stores voice features, and the operation unit stores the voice features and the input voice features. Calculate it. When the user enters the voiceprint verification system 1, the account number is first input to confirm the identity. Then, the system queries the database according to the input account, is it
Κ如、咖0卜|__1 —8 — (C 1270605 94. 08· 08第94119966號專利說明書及圖式修正本 =輸入帳號屬於已建立。若雜人帳號未建立時 ,進入該訓 二ίΐ10進行語音辑作業’續建立及儲存雜入帳號之 料。若㊅輸人帳號已建立時,進人該測試祕20進行 語音測試作業,以便辨識該輸人帳號之語音特徵是否符合已 儲存該輸入帳號之語音資料。Κ如,咖0卜|__1 —8 — (C 1270605 94. 08· 08 No. 94119966 Patent specification and schema revision = input account is established. If the miscellaneous account is not established, enter the training 2 ΐ 10 The voice series operation 'continued to establish and store the miscellaneous account materials. If the six input account has been established, enter the test secret 20 to conduct a voice test operation, in order to identify whether the voice feature of the input account meets the stored input account. Voice material.
在擷取騎特徵之前,糊該前端處理部將有效語音資 =自原始輸人語音資料擷取,㈣除無效語音資訊。本發明 、測包含短時距能量〔ShQr卜Energy〕及過料〔如〇_ ^os^ing Rate〕。本發明採用結合高斯機率分佈的 法,其方程式如下: =為原始訊號將其分為數個D維的音框、响,」 μ,為所屬機率、$為背景雜訊之如诊伯 . 的變異數。在·此,因A ?由,之期望值4為背景雜言 因為^ρττ中的D=256為一個定值,故將$ 省略不^計1算,,方程式⑴簡化如下:Before the riding feature is captured, the front-end processing unit extracts the effective voice resource = the original voice data, and (4) the invalid voice information. The invention includes the short-time energy [ShQr Bu Energy] and the over-feed [such as 〇 _ ^ os ^ing Rate]. The invention adopts a method combining the Gaussian probability distribution, and the equation is as follows: = the original signal is divided into several D-dimensional sound boxes, ringing, "μ", which is the probability of the adjacent probability, and the background noise is the diagnosis. number. In this case, because of A?, the expected value of 4 is the background. Because D=256 in ^ρττ is a fixed value, the value of $ is omitted, and equation (1) is simplified as follows:
^W = ^exp{_l(i_-)V(-__| (2) h (ύΛ \^ί 1 Γ 1 ί)) (3) 零率256點,計算短時距能量及 代入該方程式個數及原輸入語音資 丁運^。利用短時距能量與過零率的分 以⑽《糾讲咖卜《_醜3·奴 4 1270605 94· 08· 08第94119966號專利說明書及圖式修正本 機率區分有效語音資訊及無效語音資訊,將無效語音資訊加 以濾除,不但減少資料量,亦能正確擷取有效語音資訊、。 _在該特徵擷取部進行擷取特徵上,本發明採用兩個語音 硪別特徵參數,其包含線性預測倒頻譜係數〔。此虹 ^ediction Cepstrum c〇efficient,Lpa〕及梅爾頻標倒頻 谱參數〔Mel Frequency Cepstrum Coefficient,MFCC〕兩 者各12個倒頻譜參數(cepstral c〇eff沁丨印匕)及個一階 J頻4參數(delta-cepstral coefficients)。將倒頻譜參數 cn對時間倣偏微分 乂^W = ^exp{_l(i_-)V(-__| (2) h (ύΛ \^ί 1 Γ 1 ί)) (3) Zero rate 256 points, calculate the short-term energy and substitute the number of equations And the original input voice 丁丁运^. Using the short-term energy and the zero-crossing rate to divide the effective voice information and the invalid voice information by (10) "The Talk of the Talk" (_ ugly 3 · slave 4 1270605 94 · 08 · 08 94119966 patent specification and schema correction of the local rate , filter out invalid voice information, not only reduce the amount of data, but also correctly capture valid voice information. In the feature extraction section, the present invention employs two speech discriminant feature parameters including linear predictive cepstral coefficients. This rainbow ^ediction Cepstrum c〇efficient, Lpa] and Mel Frequency Cepstrum Coefficient (MFCC) are 12 cepstral parameters (cepstral c〇eff沁丨) and first order J-frequency 4 parameters (delta-cepstral coefficients). The cepstrum parameter cn is differentially differentiated from time 乂
K ㈣) dt ^- (4) Σ灸2K (4)) dt ^- (4) Acupuncture 2
k念一 K y K為考慮音框數。 ’Ί,ϋ為-階倒賴參數的公式⑷過於複雜,故將其加以 簡化,下列各式域考慮祕各兩㈣框時,絲式簡化如 下·· △C”°=[2*C(2,《) + C(1,n)]/5 (5) AC:=[2*C(35n) + C(2,n)^C(05n)]/6 ⑹ △cr[2 C(z + 2,n) + C(i + 1,n) —印―w —2*c(i —2,州削⑺ c(I—3n)—2*c(i—4,叫/6 ⑻ 1 =[—C(/:K2*C(L-3,”)]/5 ⑼ 方程式(5)至(9)中,Cn為n階特徵值,L為訊號中時框 總數,i為時框編號。 第2圖揭不本發明較佳實施例聲紋密碼鎖系統所採用的 雜驗證純之狀態及音框之隱示意圖。 在訓束m曰上,語音具有所謂「狀態」的觀念,狀態是 C:\Linda\PK Pat\PK9501-10000\PK98l3. doc η5/08/ί)5/Ι2;ηΐ in —10 — 1270605 94. 08. 08第94119966號專利說明書及囷式修正本 發音時嘴型以及聲道的變化。一般而言,每一次說話嘴型一 定有變化,故每一個狀態都是一個語音變化的特徵表現。有 時一個單音卻有可能含有多個狀態。一個狀態並不像音框一 樣具有固定尺寸,通常一個狀態包含數個或數十個音框。 ,請參照第2圖所示,第一狀態包含三個音框、第二狀態 包含六個音框及第三狀態包含四個音框。本發明首先利用均 切方式進行假定該狀態與音框之間關係。接著,本發明計算 初始模型的參數,其包含期望值及變異數〕。利用該初始模 # 型的參數重新分配音框與狀態之間關係,以獲得新切點。再 利用該新切點重新計算每個狀態所對應的音框。一再重複重 新刀配曰框與狀態之間關係及重新計算每個狀態所對應的音 框,直至最大相似機率不再上升為止。 帛3圖揭示本發明較佳實施例聲紋密碼⑽統所採用的 聲紋驗證系統之音框與狀態之初始分配模式示意圖。該初始 分配模式舉例三個樣本語音進行均分動作。 广ίΓΓ模式將語音料分動作,在均分後可能無法整 ;f平分在第—個及最後—個狀態°請再參 1、、第:二在分配模式中,樣本語音均分必須考慮三點: 於第—錄態;2、最後_個音框一定 個狀13、音框的狀態變化,只有不變或變化至 率,並且輯料料個音_於每錄態的機 革並且利用維特比演算法獲得最相似路徑。 聲紋驗第證例聲紋密—的 睛參照第4圖所示,在三個_時,l個音框可能狀態 C:\Linde\PK P*t\PK950l-|〇〇〇〇\p^98i3. ^ 05/08/05/12.01 ID y,·.....k Nian K y K is to consider the number of frames. The formula (4) of 'Ί, ϋ--------------------------------------------------------------------------- 2, ") + C(1,n)]/5 (5) AC:=[2*C(35n) + C(2,n)^C(05n)]/6 (6) △cr[2 C(z + 2,n) + C(i + 1,n) —印—w —2*c(i —2, state cut (7) c(I—3n)—2*c(i—4, call /6 (8) 1 =[—C(/:K2*C(L-3,”)]/5 (9) In equations (5) to (9), Cn is the nth-order eigenvalue, L is the total number of frames in the signal, and i is the time frame. Figure 2 is a schematic diagram showing the state of miscellaneous verification and the hidden frame of the sound frame used in the voiceprint code lock system of the preferred embodiment of the present invention. On the training m曰, the voice has the concept of "state", state It is C:\Linda\PK Pat\PK9501-10000\PK98l3. doc η5/08/ί)5/Ι2; ηΐ in —10 — 1270605 94. 08. 08 Patent No. 94191996 and the modified version of the mouth when pronounced Type and channel changes. Generally speaking, each time the mouth shape must change, each state is a characteristic of a phonetic change. Sometimes a single tone may have multiple states. A state does not have a fixed size like a sound box. Usually a state contains several or dozens of sound boxes. Please refer to Figure 2, the first state contains three sound boxes, and the second state contains six sound boxes. The sound box and the third state comprise four sound boxes. The present invention firstly uses the slashing method to assume the relationship between the state and the sound box. Next, the present invention calculates the parameters of the initial model, which include the expected value and the variation number. The parameters of the initial model # redistribute the relationship between the frame and the state to obtain a new point. Then use the new point to recalculate the frame corresponding to each state. Repeat the relationship between the frame and the state again and again. Recalculating the sound box corresponding to each state until the maximum similar probability no longer rises. Figure 3 is a diagram showing the initial allocation of the sound frame and state of the voiceprint verification system used in the voiceprint password (10) system of the preferred embodiment of the present invention. Schematic diagram of the mode. The initial allocation mode is an example of three sample voices for equalizing actions. The wide mode of the voice divides the voice material into actions, which may not be possible after the equalization; And the last state - please refer to 1, and the second: in the distribution mode, the sample speech equalization must consider three points: in the first - recording; 2, the last _ a certain frame 13, the sound box The state change, only the constant or change to the rate, and the material is expected to be _ in each recorded state of the machine and use the Viterbi algorithm to obtain the most similar path. The sound pattern test As shown in Fig. 4, in three _, the l possible states of the sound box are C:\Linde\PK P*t\PK950l-|〇〇〇〇\p^98i3. ^ 05/08/05/12.01 ID y ,·.....
A ~ 11 — 1270605 94 08. 08第94119966號專利說明書及圖式修正本 轉換之方向。將打又音框視為不可能屬於的狀態,箭頭的方 向視為可能狀態變化路徑。 第5圖揭示本發明較佳實施例聲紋密碼鎖系統所採用的 聲紋驗證系統之最相似路徑示意圖。請參照第5圖所示,擷 取特徵之最相似路徑具有第一狀態包含第i至3音框、第二 狀態包含第4至6音框及第三狀態包含第7至1〇音框。 ^第6圖揭示本發明較佳實施例聲紋密碼鎖系統所採用的 聲紋驗證系狀均分音框示顏。請參照第6圖所示,三個A ~ 11 — 1270605 94 08. 08 Patent specification No. 94119966 and schema revision The direction of conversion. The box is considered to be an impossible state, and the direction of the arrow is regarded as a possible state change path. Figure 5 is a diagram showing the most similar path of the voiceprint verification system used in the voiceprint lock system of the preferred embodiment of the present invention. Referring to Figure 5, the most similar path of the feature has a first state comprising the ith to 3rd frames, a second state comprising the 4th to 6th frames, and a third state comprising the 7th to 1st frames. Fig. 6 is a view showing the voiceprint verification system for the voiceprint combination lock system of the preferred embodiment of the present invention. Please refer to Figure 6, three
樣本語音在三個狀態的初始模式,其採均分後之分佈。第一 樣本語音之每個樣本語音均分三個音框後,剩餘兩個音框分 別刀配置帛狀悲及第二狀態。第二樣本語音之每個樣本語 音均分四個音框。第三樣本語音之每個樣本語音均分三個音 框後,剩餘-個音框分別分配置第—狀態。在計算後,盆二 大相似機率為2156。 ,、取 /第7圖揭示本發明較佳實施例聲紋密碼鎖系統所 策、文驗d统之第—次重新分配音框示意圖。請參照第7圖 =。’在第—次重新分配音域,其最幼似機率上升: 所示 3571 〇 第8圖揭示本發明較佳實施例聲纹密所 秋驗證纽之第二轉新分配雜示意圖。料照第= 在第一次重新分配音減,其最大相似機率上升至 第9圖揭示本發明較佳實施例聲紋 =驗證系統之最佳分配音框示意圖。請參照第== 在夕-人重新触音框後,其最大相似機率3571不再上升,因 C:\LimU\PK Pftt、PK95〇M_)fl\PKMI3· doc —12 — 〇5/〇8/〇5/|2;〇| •1270605 08第94119966號專利說明書及圖式修正本 的· υο· 两久闽八修正本 =其,為最佳分配音框。計算各狀態的射 檔可供儲存在^庫成极讀標,將該聲紋特徵 請再參照第1圖所示,在進入該訓練系統10進的立 者1用維特比演算法獲得最她雜 狀 == 為模_,即完成語二= 率:於-預定參考值時,無法通過 統! ____聲紋驗證系 立餐 八最大相似機率大於該預定參考值時,通過往 行加密〔~^ 再Γ第1圖所示,同樣的在進入該測試系統2〇進 二1曰二作業時’運算方程式⑴至⑼獲得有效測試語音 時’將該聲紋特徵槽進行解密〔〕,以 Μ聲紋驗證系統1進行聲紋驗證作業。 命没/再/:第1圖所不’接著,進行運算該測試語音特徵 …尸^寺欲檔之模型參數之間相似機率以便輸出一辨識結 果:在語音觸上,其最小她機社於就參考值時,通 過叩曰因而可進人該聲紋驗證系統1 ;反之,其最小 相似機率小於該縱參考_,無㈣過語音_且結束測 试作業,因而必須離開該聲紋驗證系統卜 ,人〇圖揭示本發明第-較佳實施例聲紋密碼鎖系統利 用固疋式雜密碼鎖進行加密及解密之流程方制。請參照 C:\Linde\PK ft«t\PM50M⑽mpweiw 05/08/05/12:01 fThe sample speech is in the initial mode of three states, and it is distributed after the equalization. After each sample voice of the first sample speech is divided into three sound boxes, the remaining two sound boxes are respectively arranged in a shape and a second state. Each sample of the second sample speech is divided into four frames. After each sample of the third sample speech is divided into three frames, the remaining - frames are respectively assigned the first state. After calculation, the probability of the second largest similarity is 2156. FIG. 7 is a schematic diagram showing the first-time redistribution sound box of the voiceprint code lock system according to the preferred embodiment of the present invention. Please refer to Figure 7 =. In the first redistribution of the sound field, its youngest chance is increased: 3571 所示 Figure 8 shows a schematic diagram of the second transfer new distribution of the sound-printing secrets of the preferred embodiment of the present invention. The photographing = the first redistribution of the sound reduction, the maximum similar probability of which rises to Fig. 9 reveals a schematic diagram of the best distribution sound frame of the voiceprint = verification system of the preferred embodiment of the invention. Please refer to the === after the eve-person re-telephone, its maximum similar probability 3571 no longer rises, because C:\LimU\PK Pftt, PK95〇M_)fl\PKMI3· doc —12 — 〇5/〇8 〇 / / • • • • • • • • • • • • • • • • • • • • • • • • 941 941 941 941 941 941 941 941 941 941 941 941 941 941 941 941 941 941 941 941 The calculation of the projections of each state can be stored in the library. The voiceprint features can be further referred to in Figure 1, and the first person entering the training system 10 uses the Viterbi algorithm to obtain the most. Miscellaneous == is the modulo _, that is, the completion of the second = rate: at - predetermined reference value, can not pass the system! ____ voiceprint verification system 8 when the maximum similar probability is greater than the predetermined reference value, encrypted by the line [~^ again, as shown in Figure 1, the same when entering the test system 2 into the 2nd and 2nd work 'Operation equations (1) to (9) 'decrypt the voiceprint feature groove when the valid test voice is obtained, and perform the voiceprint verification operation with the voiceprint verification system 1. Life does not / again /: Figure 1 does not 'Next, the operation of the test voice features ... corpse ^ Temple wants the file parameters between the similar probability to output a recognition result: in the voice touch, the smallest her machine As far as the reference value is concerned, the voiceprint verification system 1 can be entered by 叩曰; otherwise, the minimum similarity probability is less than the vertical reference _, no (four) over-speech_ and the test operation is ended, and thus the voiceprint verification system must be left. In view of the above, the voiceprint lock system of the first preferred embodiment of the present invention utilizes a solid-type miscellaneous code lock for encryption and decryption. Please refer to C:\Linde\PK ft«t\PM50M(10)mpweiw 05/08/05/12:01 f
Cr) —13 1270605 94 08. 08第94119966號專利說明書及圖式修正本 第圖所示,本發明第一較佳實施例聲紋密碼鎖系統3係屬 固定式聲紋密碼鎖,其包含__聲紋餘匙Ke。該聲紋餘匙^ 用=將-電子資料進行加密或解密運算,以構成該固定式聲 紋毯碼鎖。該固定式聲紋密碼鎖適用於個人電腦〔π〕、筆 記型電腦〔NB〕、個人數位助理〔PDA〕或行動電話〔M〇biie Phone〕等。 < 士請再參照第1及10 @所示’本發明聲紋密碼鎖系統: 採用聲紋驗證系統i之訓練系統10。該聲紋驗證系統以 產生=聲_徵值’自該較雜值選取該聲⑽匙^。該 訓練系統10可提供一聲紋特徵檔31,其較佳係由該聲紋特 徵檔操取32個位元組〔byte〕。所謂聲紋餘匙係電子資料的 儲存端及摘取端皆需要輸入相同的聲紋特徵。在儲存電子資 I 1 10 Ϊ獲聲输匙Kc。此時,該聲紋餘起Kc用以將-電子 貝=播32進行加密縣,並將6加密電子資料檔儲存在特定 3 ^卩完成電子靜^密作業。本發賴佳制先進加密 δΜΓ(1 5 AES) 明再參照第1及1〇圖所示,在擷 將統1之測試系統20進行語音測試作業,以便 與輸入聲音進行比對。若該聲紋特刪 二旦通過i音二=;;預設值時,視其通過語音測試作業。 反之,若轉:轉$,即可允許開啟該電子資料播32。 以表次特_ 31與輸人聲音之誤差高於—預設值 C:\Linda\PK Ρ*1\ΡΚ950 Μ0000\ρΚ9ίΐ3. doc 05/08/05/12:01 Pll ~~~ 14 —— 1270605 ^ 94 08· 08第94119966號專利說明書及囷式修正本 時’視其未通過語音測試作業。一旦未通過語音測試作業時, =需離開該聲紋密碼鎖系統3 1不允許·該電子資料檔 、\1圖揭不本發明第二較佳實施例聲紋密碼鎖系統利 “ 用可攜絲紋密碼鎖進行加密及解密之流财塊圖。請來昭 第11圖所示,本發明第二較佳實施例聲紋密碼鎖系統4係屬、 3攜式聲紋密碼鎖。該聲紋鑰匙KG用以將—電子資料進行加 密或解密,輯賴可攜式聲㈣補。該可攜式 ❿密碼鎖適用於光碟〔⑻、軟碟〔Floppy Disk〕、隨身碟 〔Flash Disk〕、磁光碟〔mo Disk〕或網路傳輸等。 再參照第1及11圖所示,本發明聲紋密碼鎖系統4 _ 採用聲紋驗證系統1之訓練系統10,該訓練系統10可提供 一聲紋特徵檔41。該聲紋驗證系統丨用以產生一 自該聲紋特徵值縣鱗紋鑰匙KG。在儲存電子值可 ,利肋聲紋驗證系統丨之鱗系統1()之輯作業獲得該 聲紋鍮匙Kc。此時,將該聲紋特徵播41 *入該電子資料," • 即該聲紋特徵檔41僅占該電子資料檔42的2K至61(二位元 組。利用該聲紋鑰匙Kc將該電子資料檔42進行加密運算兀 以產生加密電子資料槽;同時顧—般金餘κ將該^文二徵 ^ 41亦進行加密運算,以產生加密聲紋特徵播,並將該加密 聲紋特徵檔及加密電子資料檔加以串聯,以獲得一串聯檔。 另外,將該加密聲紋特徵檔及加密電子資料檔之串聯播^〇經 訊息認證碼產生機制產生訊息認證碼。本發明較佳採曰用安全 雜散演算法〔Secure Hash Algorithm,簡稱sha〕產生訊φ 認證碼,並利用一般金鑰K將該訊息認證碼進行加密運算了 —15 — C:\Linda\PK ΡΜ\ΡΚ95(Μ-|〇〇〇〇\ρΚ9β|3 知Cr) - 13 1270605 94 08. 08 Patent No. 94191996 and the drawings are modified. The first preferred embodiment of the present invention is a fixed voiceprint lock system, which includes _ _ voice relief key Ke. The voiceprint key ^ is used to encrypt or decrypt the electronic data to form the fixed voice blanket code lock. The fixed voiceprint combination lock is suitable for a personal computer [π], a notebook computer (NB), a personal digital assistant (PDA), or a mobile phone (M〇biie Phone). < Further, please refer to the first and tenth @@'s voiceprint lock system of the present invention: The training system 10 using the voiceprint verification system i. The voiceprint verification system selects the sound (10) key ^ from the more complicated value by generating a = sound_signal value. The training system 10 can provide a voiceprint profile 31, which preferably operates 32 bytes from the voiceprint feature. The so-called voiceprint features are required to input the same voiceprint feature on both the storage end and the pick-up end of the electronic data. In the storage of electronic funds I 1 10 capture the sound input key Kc. At this time, the voiceprint remaining Kc is used to encrypt the county, and the 6-encrypted electronic data file is stored in a specific 3^卩 electronic electronic security operation. This is the advanced encryption of the Laijia system. δΜΓ(1 5 AES) As shown in the first and the first diagrams, the test system 20 of the system 1 is subjected to a voice test operation to compare with the input sound. If the voiceprint is specially deleted, the second tone is passed through the i tone two =;; when the preset value is used, it is regarded as a voice test operation. Conversely, if you turn: turn $, you can open the electronic data broadcast 32. The error between the table _ 31 and the input voice is higher than the default value C:\Linda\PK Ρ*1\ΡΚ950 Μ0000\ρΚ9ίΐ3. doc 05/08/05/12:01 Pll ~~~ 14 —— 1270605 ^ 94 08· 08 No. 94119966 Patent Specification and Modifications are considered to have failed to pass the voice test. Once the voice test operation is not passed, = need to leave the voiceprint password lock system 3 1 is not allowed. The electronic data file, \1 Figure does not reveal the second preferred embodiment of the present invention, the voiceprint password lock system is "portable The flow code block diagram of the silk code lock is used for encryption and decryption. As shown in Fig. 11, the voiceprint lock system 4 of the second preferred embodiment of the present invention is a three-sound voice code lock. The KG is used to encrypt or decrypt electronic data. It is compatible with portable audio (4). The portable security code lock is suitable for CDs (8), Floppy Disk, Flash Disk, Mo Disk or network transmission, etc. Referring again to Figures 1 and 11, the voiceprint lock system 4 of the present invention uses the training system 10 of the voiceprint verification system 1, and the training system 10 provides a sound. The pattern feature file 41. The voiceprint verification system is used to generate a scale key KG from the voiceprint feature value. In the storage of the electronic value, the scale calculation system of the Rib voiceprint verification system is obtained. The voiceprint key Kc. At this time, the voiceprint feature is broadcasted into the electronic data, & That is, the voiceprint profile 41 only occupies 2K to 61 (two bytes) of the electronic data file 42. The electronic data file 42 is encrypted by the voiceprint key Kc to generate an encrypted electronic data slot; At the same time, the grammar operation is performed to generate an encrypted voiceprint feature, and the encrypted voiceprint profile and the encrypted electronic data file are serially connected to obtain a serial file. In addition, the encrypted voiceprint profile and the encrypted electronic data file are serially transmitted to generate a message authentication code by using a message authentication code generation mechanism. The present invention preferably uses a Secure Hash Algorithm (sha). Generate the signal φ authentication code, and use the general key K to encrypt the message authentication code—15—C:\Linda\PK ΡΜ\ΡΚ95(Μ-|〇〇〇〇\ρΚ9β|3
05/08/05/12:01 IH 1270605 94· 〇8· 08第94119966號專利說明書及圖式修正本 以產生加密訊息認證碼。該加密電子資繼、加密聲紋特徵 檔及加密訊息認證碼形成—可攜式齡,即完子資料加 密作業,其可經由網路或記錄媒體傳送。 is? 明再參照第1及11圖所示,在—賴系統擷取電子資 料時’首先將該加密聲紋特徵檔及加密訊息認證碼利用一般 金鑰κ進行解密運算,峨得鱗_漏41及訊息認證碼。 接著,利用該聲紋驗證系統i之測試系統2〇進行語音測試作 業,以便將該聲紋特_ 41與輸人聲音進行比對。若該聲紋 特徵檔41與輸人聲音之誤差餘—職_,視其通過語音 測試作業…旦通過語音峨健時,即可允許由該聲 匙^開啟該電子資料檔42。反之,若該聲紋特徵播Μ與^ 入聲曰之―綠1設值時,視其未通聰音職作業。 -旦未通過語音職作料,即需離_聲紋 统4, 且不允許開啟該電子資料檔42。 請再參照第11圖所示,最後,將該聲紋特徵槽41 進自行認證碼比對。若嫩特徵檔 及電子貝仙42付合訊息認證碼時,該電子資料標犯可 顯不於該電«統。反之,若鱗__ 41及電 42不符合訊息認證碼時’該電子資料槽犯|法顯示於= 腦系統,且需離開該聲紋密碼鎖系統4。 .、以電 雖然本發明已以前述較佳實施例揭示,麸i 定本發明,任何熟習此技藝者,在 以限 圍内’當可作各種之更動與修改,因 視後附之申請專麵_界定者鱗。 4乾圍虽 —16 — C:\Linde\PK Pht\PK950MOOOO\PIC9ei3. doc05/08/05/12:01 IH 1270605 94· 〇8· 08 No. 94119966 Patent Specification and Schema Revision to generate an encrypted message authentication code. The encrypted electronic asset, the encrypted voiceprint profile, and the encrypted message authentication code are formed—portable age, that is, the encrypted data encryption operation, which can be transmitted via the network or the recording medium. Referring to Figures 1 and 11, as shown in Figure 1 and Figure 11, when the electronic data is retrieved from the system, the encrypted voice signature and the encrypted message authentication code are first decrypted using the general key κ. 41 and message authentication code. Next, the test system 2 of the voiceprint verification system i is used to perform a voice test operation to compare the voiceprint 41 with the input voice. If the error between the voiceprint profile 41 and the input voice is __, it can be allowed to be turned on by the voice key when it passes the voice test operation. On the other hand, if the voiceprint feature is broadcasted and the "green" value of the voice is set, it is considered to be unclear. - If you have not passed the voice job, you need to leave the whispering system 4, and you are not allowed to open the electronic data file 42. Please refer to FIG. 11 again. Finally, the voiceprint feature slot 41 is compared with the self-certification code. If the tender feature file and the electronic beixiang 42 are combined with the message authentication code, the electronic data offender may not be displayed. On the other hand, if the scale __ 41 and the power 42 do not meet the message authentication code, the electronic data slot is displayed in the brain system and needs to leave the voiceprint lock system 4. Although the present invention has been disclosed in the foregoing preferred embodiments, the invention is invented by the present invention, and any person skilled in the art can make various changes and modifications within the scope of the limitation. _ define the scales. 4 dry circumference though —16 — C:\Linde\PK Pht\PK950MOOOO\PIC9ei3. doc
n5/0g/05/l2;fl| PM 1270605 94· 08. 08第94119966號專利說明書及囷式修正本 【圖式簡單說明】 第1圖:本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之流程方塊圖。 第2圖·本發明實施例聲紋密碼鎖糸統所採用的聲紋驗證系 統之狀態及音框之關係示意圖。 第3圖:本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之音框與狀態之初始分配模式示意圖。 f 第4圖:本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之狀態轉換示意圖。 第5圖:本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之最相似路徑示意圖。 第6圖:本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之均分音框示意圖。 第7圖··本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之第一次重新分配音框示意圖。 _ 第8圖:本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之第二次重新分配音框示意圖。 第9圖··本發明實施例聲紋密碼鎖系統所採用的聲紋驗證系 統之最佳分配音框示意圖。 第10圖:本發明第一較佳實施例聲紋密碼鎖系統利用固定 式聲紋密碼鎖進行加密及解密之流程方塊圖。 第11圖:本發明第二較佳實施例聲紋密碼鎖系統利用可攜 式聲紋密碼鎖進行加密及解密之流程方塊圖。 C:\Ltnda\PI PKt\PK950M0000\PK98l3. doc —17 — ⑧ 1270605 94. 08. 08第94119966號專利說明書及圖式修正本 【主要元件符號說明】 1 聲紋驗證系統 20 測試系統 31 聲紋特徵檔 4 聲紋密碼鎖系統 41 聲紋特徵檔 10訓練系統 3 聲紋密碼鎖系統 32電子資料檔 40串聯檔 42電子資料檔N5/0g/05/l2;fl| PM 1270605 94· 08. 08 Patent No. 94119966 and 囷 修正 【 【 简单 简单 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第Process block diagram of the pattern verification system. Fig. 2 is a diagram showing the relationship between the state of the voiceprint verification system and the sound frame used in the voiceprint code lock system according to the embodiment of the present invention. Fig. 3 is a schematic view showing the initial allocation pattern of the sound frame and state of the voiceprint verification system used in the voiceprint combination lock system of the embodiment of the present invention. f Fig. 4 is a schematic diagram showing the state transition of the voiceprint verification system used in the voiceprint lock system of the embodiment of the present invention. Fig. 5 is a view showing the most similar path of the voiceprint verification system used in the voiceprint lock system of the embodiment of the present invention. Fig. 6 is a schematic diagram showing the sound distribution frame of the voiceprint verification system used in the voiceprint code lock system of the embodiment of the present invention. Fig. 7 is a schematic view showing the first redistribution of the voiceprint verification system used in the voiceprint code lock system of the embodiment of the present invention. _ Figure 8 is a schematic diagram of the second redistribution sound box of the voiceprint verification system used in the voiceprint lock system of the embodiment of the present invention. Fig. 9 is a schematic view showing the best sound distribution frame of the voiceprint verification system used in the voiceprint lock system of the embodiment of the present invention. Fig. 10 is a block diagram showing the flow of encryption and decryption using a fixed voiceprint combination lock according to a first preferred embodiment of the present invention. Figure 11 is a block diagram showing the flow of encryption and decryption using a portable voiceprint combination lock according to a second preferred embodiment of the present invention. C:\Ltnda\PI PKt\PK950M0000\PK98l3. doc —17 — 8 1270605 94. 08. 08 Patent No. 94191996 and schema revision [Main component symbol description] 1 Voiceprint verification system 20 Test system 31 Voiceprint Feature file 4 voiceprint password lock system 41 voiceprint feature file 10 training system 3 voiceprint password lock system 32 electronic data file 40 serial file 42 electronic data file
Claims (1)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW94119966A TWI270605B (en) | 2005-06-16 | 2005-06-16 | Voiceprint key system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW94119966A TWI270605B (en) | 2005-06-16 | 2005-06-16 | Voiceprint key system |
Publications (2)
Publication Number | Publication Date |
---|---|
TW200700630A TW200700630A (en) | 2007-01-01 |
TWI270605B true TWI270605B (en) | 2007-01-11 |
Family
ID=38430191
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW94119966A TWI270605B (en) | 2005-06-16 | 2005-06-16 | Voiceprint key system |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI270605B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI481774B (en) * | 2013-09-18 | 2015-04-21 | Generalplus Technology Inc | Method for unlocking door, method for leasing asset and system thereof |
-
2005
- 2005-06-16 TW TW94119966A patent/TWI270605B/en not_active IP Right Cessation
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI481774B (en) * | 2013-09-18 | 2015-04-21 | Generalplus Technology Inc | Method for unlocking door, method for leasing asset and system thereof |
Also Published As
Publication number | Publication date |
---|---|
TW200700630A (en) | 2007-01-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR101224348B1 (en) | Personal communication apparatus capable of recording transactions secured with biometric data, and computer readable recording medium | |
US9430628B2 (en) | Access authorization based on synthetic biometric data and non-biometric data | |
US9674184B2 (en) | Systems and methods to generate authorization data based on biometric data and non-biometric data | |
CN103391201B (en) | The system and method for smart card identity checking is realized based on Application on Voiceprint Recognition | |
CN110659468B (en) | File encryption and decryption system based on C/S architecture and speaker identification technology | |
TW200923798A (en) | Biometrics method and apparatus and biometric data encryption method thereof | |
Wang et al. | Toothsonic: Earable authentication via acoustic toothprint | |
US20070038868A1 (en) | Voiceprint-lock system for electronic data | |
JP4351659B2 (en) | Voiceprint password key system | |
Nagakrishnan et al. | A robust cryptosystem to enhance the security in speech based person authentication | |
ES2870153T3 (en) | Method and system for creating an electronic signature of a document associated with a person by the person's voice print and corresponding method to verify the electronic signature | |
Zhang et al. | Voicepm: A robust privacy measurement on voice anonymity | |
TWI270605B (en) | Voiceprint key system | |
Wang et al. | Earable authentication via acoustic toothprint | |
JP2007220068A (en) | Portable voiceprint password key remote transmission system and method for it | |
EP1760566A1 (en) | Voiceprint-lock system for electronic data | |
CN100444188C (en) | Vocal-print puzzle lock system | |
CN106682531A (en) | Method for confidential data encryption based on biological information authorization | |
Sasikaladevi et al. | SCAN-speech biometric template protection based on genus-2 hyper elliptic curve | |
Mtibaa | Towards robust and privacy-preserving speaker verification systems | |
Johnson | Privacy enhanced remote voice verification | |
de Groot | Biometric security on body sensor networks | |
Inthavisas | Secure speech biometric templates | |
TW200816166A (en) | Randomized voiceprint identification system, random voiceprint lock and creating method therefor | |
SQUARTINI et al. | 284 Neural Nets WIRN10 B. Apolloni et al.(Eds.) IOS Press, 2011© 2011 The authors and IOS Press. All rights reserved. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
MM4A | Annulment or lapse of patent due to non-payment of fees |