TW202328951A - Low-noise voiceprint identification device for financial transaction system and method thereof - Google Patents
Low-noise voiceprint identification device for financial transaction system and method thereof Download PDFInfo
- Publication number
- TW202328951A TW202328951A TW112103776A TW112103776A TW202328951A TW 202328951 A TW202328951 A TW 202328951A TW 112103776 A TW112103776 A TW 112103776A TW 112103776 A TW112103776 A TW 112103776A TW 202328951 A TW202328951 A TW 202328951A
- Authority
- TW
- Taiwan
- Prior art keywords
- voiceprint
- feature
- feature data
- data
- verification
- Prior art date
Links
Images
Landscapes
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
Description
本發明係有關一種聲紋辨識裝置與其方法,尤其是有關一種用於金融交易系統之低雜訊聲紋辨識裝置與其方法。The present invention relates to a voiceprint recognition device and its method, in particular to a low-noise voiceprint recognition device and its method for a financial transaction system.
聲紋具有唯一性、獨特性和不易竄改性,加之較不涉及使用者隱私問題,故聲紋識別可用於要求從音頻信號中提取個體差異,擷取出能夠反映使用者是誰的信息,從而進行使用者識別,其基本原理是每一個使用者建立一個能夠描述這一使用者個性特徵的模組,作為此使用者個性特徵的描述。Voiceprint is unique, unique and not easy to tamper with, and it does not involve user privacy issues, so voiceprint recognition can be used to extract individual differences from audio signals and extract information that can reflect who the user is, so as to carry out The basic principle of user identification is that each user creates a module that can describe the user's personality characteristics as a description of the user's personality characteristics.
聲紋辨識目前已經應用於身份辨識,作為判斷使用者身份之工具。然而,聲紋識別的缺點在於,聲音容易受到年齡、情緒、或身體狀況等因素影響,導致識別性能降低,故目前聲紋辨識技術的準確度仍有待改善。此外設備的收音品質與環境噪音等因素,也會影響聲紋辨識的識別效能。Voiceprint recognition has been applied to identity recognition as a tool for judging the user's identity. However, the disadvantage of voiceprint recognition is that the voice is easily affected by factors such as age, emotion, or physical condition, resulting in reduced recognition performance. Therefore, the accuracy of the current voiceprint recognition technology still needs to be improved. In addition, factors such as the sound quality of the device and environmental noise will also affect the recognition performance of voiceprint recognition.
有鑑於先前技術的上述缺點,本發明之一態樣提供了一低雜訊聲紋辨識裝置,該裝置包含:一音訊接收單元,係用以擷取一音頻資料;一聲紋辨識單元,係連接於該音訊接收單元,該聲紋辨識單元由該音訊接收單元接收該音頻資料,且該聲紋辨識單元辨識出該音頻資料的至少一聲紋特徵;一儲存單元,儲存一第一特徵資料;一運算單元,係連接於該聲紋辨識單元與該儲存單元,該運算單元自該聲紋辨識單元接收該至少一聲紋特徵以及自該儲存單元接收該第一特徵資料,且該運算單元包含:一驗證計算器,該驗證計算器之計數值於一初始狀態為0單位,其中該運算單元根據該第一特徵資料判斷該至少一聲紋特徵與該第一特徵資料的是否相符;其中,若該運算單元判斷該至少一聲紋特徵符合該第一特徵資料,則該運算單元將該至少一聲紋特徵儲存於該儲存單元的一暫存空間內,根據該暫存空間內的該至少一聲紋特徵訓練該第一特徵資料,以產生一調整後第一特徵資料,並根據該調整後第一特徵資料進行安全性驗證;若該運算單元判斷該至少一聲紋特徵不符該第一特徵資料,則該運算單元傳送一驗證請求,並於該驗證請求完成確認後,該運算單元將該至少一聲紋特徵儲存於該暫存空間內,但該儲存單元不根據該至少一聲紋特徵變更該第一特徵資料,以根據該第一特徵資料進行安全性驗證,其中該驗證請求包含使用者身份特徵之識別;若該運算單元判斷該至少一聲紋特徵不符該第一特徵資料,且該計算值達該預定數值,則該運算單元根據該暫存空間內的該至少一聲紋特徵與該第一特徵資料產生一第二特徵資料;以及於該第二特徵資料產生後,該驗證計算器調整該計算值為0單位,該運算單元將該第一特徵資料、該第二特徵資料儲存於儲存單元,以根據該第一特徵資料、該第二特徵資料判斷該至少一聲紋特徵與該第一特徵資料、該第二特徵資料是否相符,以進行安全性驗證,包含:若該運算單元判斷該至少一聲紋特徵符合該第二特徵資料,則該運算單元傳送該至少一聲紋特徵儲存於該暫存空間內,且根據該暫存空間內的該至少一聲紋特徵訓練該第二特徵資料,以產生一調整後第二特徵資料,並根據該調整後第二特徵資料進行安全性驗證;若該運算單元判斷該至少一聲紋特徵不符該第二特徵資料、但符合該第一特徵資料,則該運算單元自該儲存單元移除該第二特徵資料,以根據該第一特徵資料進行安全性驗證;以及於該第二特徵資料產生後,若該運算單元判斷該至少一聲紋特徵不符該第二特徵資料且不符該第一特徵資料,則該驗證計算器對該計數值增加1單位,且當該計算值小於一預定數值時,該運算單元傳送一驗證請求,並於該驗證請求完成確認後,該運算單元將該至少一聲紋特徵儲存於該暫存空間內,但該儲存單元不根據該至少一聲紋特徵變更該第一特徵資料、該第二特徵資料,以根據該第一特徵資料、該第二特徵資料進行安全性驗證,其中辨識該音頻資料內的該至少一聲紋特徵進一步包含以下操作中至少一者:濾波、降低雜訊、抑制背景噪音、放大特定聲紋特徵、計算該音頻資料之向量參數、偵測特定關鍵詞、偵測特定聲波波段、偵測特定聲波波形以及偵測特定聲波頻率。In view of the above-mentioned shortcomings of the prior art, an aspect of the present invention provides a low-noise voiceprint recognition device, which includes: an audio receiving unit for retrieving an audio data; a voiceprint recognition unit for connected to the audio receiving unit, the voiceprint recognition unit receives the audio data from the audio receiving unit, and the voiceprint recognition unit recognizes at least the voiceprint characteristics of the audio data; a storage unit stores a first characteristic data ; a computing unit connected to the voiceprint recognition unit and the storage unit, the computing unit receives the at least voiceprint features from the voiceprint recognition unit and the first characteristic data from the storage unit, and the computing unit Including: a verification calculator, the count value of the verification calculator is 0 units in an initial state, wherein the operation unit judges whether the at least voiceprint feature matches the first feature data according to the first feature data; wherein , if the computing unit judges that the at least voiceprint features conform to the first feature data, then the computing unit stores the at least voiceprint features in a temporary storage space of the storage unit, according to the At least the voiceprint feature trains the first feature data to generate an adjusted first feature data, and performs security verification according to the adjusted first feature data; if the computing unit judges that the at least voiceprint feature does not match the first feature data feature data, the computing unit sends a verification request, and after the verification request is confirmed, the computing unit stores the at least one voiceprint feature in the temporary storage space, but the storage unit does not The fingerprint feature changes the first feature data to perform security verification based on the first feature data, wherein the verification request includes the identification of the user's identity feature; if the computing unit judges that the at least one voiceprint feature does not match the first feature data , and the calculated value reaches the predetermined value, then the computing unit generates a second feature data according to the at least voiceprint feature and the first feature data in the temporary storage space; and after the second feature data is generated, The verification calculator adjusts the calculation value to 0 units, and the computing unit stores the first characteristic data and the second characteristic data in the storage unit, so as to judge the at least one sound according to the first characteristic data and the second characteristic data. Whether the fingerprint features are consistent with the first feature data and the second feature data is used for security verification, including: if the computing unit judges that the at least one voiceprint feature matches the second feature data, then the computing unit sends the at least The voiceprint feature is stored in the temporary storage space, and the second feature data is trained according to the at least voiceprint feature in the temporary storage space to generate an adjusted second feature data, and according to the adjusted second feature data for safety verification; if the computing unit judges that the at least one voiceprint feature does not match the second feature data but matches the first feature data, the computing unit removes the second feature data from the storage unit, to Perform security verification according to the first feature data; and after the second feature data is generated, if the computing unit judges that the at least voiceprint feature does not match the second feature data and does not match the first feature data, the verification calculation The counter value is increased by 1 unit, and when the calculated value is less than a predetermined value, the computing unit sends a verification request, and after the verification request is confirmed, the computing unit stores the at least voiceprint feature in the In the temporary storage space, but the storage unit does not change the first characteristic data and the second characteristic data according to the at least voiceprint characteristics, so as to perform security verification based on the first characteristic data and the second characteristic data, wherein the identification The at least one voiceprint feature in the audio data further includes at least one of the following operations: filtering, reducing noise, suppressing background noise, amplifying specific voiceprint features, calculating vector parameters of the audio data, detecting specific keywords, Detect specific sound wave bands, detect specific sound wave waveforms, and detect specific sound wave frequencies.
本發明之另一態樣提供了一低雜訊聲紋辨識方法,應用於一聲紋辨識裝置,該聲紋辨識裝置包含一運算單元以及一驗證計算器,該驗證計算器之一計數值於一初始狀態為0單位,該聲紋辨識方法包含控制該運算單元進行以下步驟:擷取一音頻資料;辨識該音頻資料內的至少一聲紋特徵;提供一第一特徵資料、一暫存空間以及一驗證請求;根據該第一特徵資料,比對該至少一聲紋特徵與該第一特徵資料是否相符;若該至少一聲紋特徵符合該第一特徵資料,則儲存該至少一聲紋特徵於該暫存空間內,並根據該暫存空間內的該至少一聲紋特徵訓練該第一特徵資料,以產生一調整後第一特徵資料,以根據該調整後第一特徵資料進行安全性驗證;若該至少一聲紋特徵不符該第一特徵資料,則傳送該驗證請求,當該驗證請求完成確認後,儲存該至少一聲紋特徵於該暫存空間內,但不根據該至少一聲紋特徵變更該第一特徵資料,以根據該第一特徵資料進行安全性驗證,其中該驗證請求包含使用者身份特徵之識別;若該計算值達該預定數值,則根據該暫存空間內的該至少一聲紋特徵與該第一特徵資料產生一第二特徵資料;以及於該第二特徵資料產生後,調整該驗證計算器之該計算值為0單位,且根據該第一特徵資料、該第二特徵資料來比對該至少一聲紋特徵與該第一特徵資料、該第二特徵資料是否相符,以進行安全性驗證,包含:若該運算單元判斷該至少一聲紋特徵符合該第二特徵資料,則該運算單元傳送該至少一聲紋特徵儲存於該暫存空間內,且根據該暫存空間內的該至少一聲紋特徵訓練該第二特徵資料,以產生一調整後第二特徵資料,並根據該調整後第二特徵資料進行安全性驗證;若該運算單元判斷該至少一聲紋特徵不符該第二特徵資料、但符合該第一特徵資料,則該運算單元自該儲存單元移除該第二特徵資料,以根據該第一特徵資料進行安全性驗證;以及於該第二特徵資料產生後,若該運算單元判斷該至少一聲紋特徵不符該第二特徵資料且不符該第一特徵資料,則該驗證計算器對該計數值增加1單位,且當該計算值小於一預定數值時,該運算單元傳送一驗證請求,並於該驗證請求完成確認後,該運算單元將該至少一聲紋特徵儲存於該暫存空間內,但該儲存單元不根據該至少一聲紋特徵變更該第一特徵資料、該第二特徵資料,以根據該第一特徵資料、該第二特徵資料進行安全性驗證,其中辨識該音頻資料內的該至少一聲紋特徵進一步包含以下操作中至少一者:濾波、降低雜訊、抑制背景噪音、放大特定聲紋特徵、計算該音頻資料之向量參數、偵測特定關鍵詞、偵測特定聲波波段、偵測特定聲波波形以及偵測特定聲波頻率。Another aspect of the present invention provides a low-noise voiceprint recognition method, which is applied to a voiceprint recognition device. The voiceprint recognition device includes an arithmetic unit and a verification calculator. One count value of the verification calculator is An initial state is 0 units, and the voiceprint recognition method includes controlling the computing unit to perform the following steps: capture an audio data; identify at least voiceprint features in the audio data; provide a first feature data, a temporary storage space and a verification request; according to the first feature data, compare whether the at least voiceprint features match the first feature data; if the at least voiceprint features match the first feature data, store the at least voiceprint features feature in the temporary storage space, and train the first feature data according to the at least one voiceprint feature in the temporary storage space to generate adjusted first feature data for performing security based on the adjusted first feature data verification; if the at least one voiceprint feature does not match the first feature data, then send the verification request, and when the verification request is confirmed, store the at least one voiceprint feature in the temporary storage space, but not according to the at least The voiceprint feature changes the first feature data to perform security verification based on the first feature data, wherein the verification request includes the identification of the user's identity feature; if the calculated value reaches the predetermined value, then according to the temporary storage space The at least one voiceprint feature and the first feature data in generate a second feature data; and after the second feature data is generated, adjust the calculation value of the verification calculator to 0 units, and according to the first feature data, the second feature data to compare whether the at least voiceprint features are consistent with the first feature data and the second feature data, so as to perform security verification, including: if the computing unit judges that the at least voiceprint features If the second feature data is met, the computing unit transmits the at least voiceprint feature to be stored in the temporary storage space, and trains the second feature data according to the at least voiceprint feature in the temporary storage space to generate a Adjusted second feature data, and perform security verification based on the adjusted second feature data; The unit removes the second feature data from the storage unit to perform security verification based on the first feature data; and after the second feature data is generated, if the computing unit determines that the at least voiceprint features do not match the second If the characteristic data does not match the first characteristic data, the verification calculator will increase the count value by 1 unit, and when the calculated value is less than a predetermined value, the operation unit will send a verification request, and after the verification request is confirmed , the computing unit stores the at least voiceprint feature in the temporary storage space, but the storage unit does not change the first feature data and the second feature data according to the at least voiceprint feature, so as to change the first feature data according to the first feature data and the second feature data for security verification, wherein identifying the at least voiceprint features in the audio data further includes at least one of the following operations: filtering, reducing noise, suppressing background noise, amplifying specific voiceprint features, Calculate vector parameters of the audio data, detect specific keywords, detect specific sound wave bands, detect specific sound wave waveforms, and detect specific sound wave frequencies.
綜上所述,使用者於登入系統環境欲進行金融交易等動作時,藉由讀取裝置使用時所記錄的聲紋特徵,搭配深度學習、加入神經網路等訓練聲紋特徵模型,可改善認證的品質,讓模型可以更準確地辨認使用者本人,亦進一步地將此項認證應用於各種金融交易系統當中。To sum up, when the user logs into the system environment and intends to perform financial transactions and other actions, by reading the voiceprint features recorded during the use of the device, and training the voiceprint feature model with deep learning and adding neural networks, it can improve The quality of authentication allows the model to more accurately identify the user himself, and further applies this authentication to various financial transaction systems.
以下說明係為完成發明的較佳實現方式,其目的在於描述本發明的基本精神,但並不用以限定本發明。實際的發明內容必須參考之後的權利要求範圍。The following description is a preferred implementation of the invention, and its purpose is to describe the basic spirit of the invention, but not to limit the invention. For the actual content of the invention, reference must be made to the scope of the claims that follow.
必須了解的是,使用於本說明書中的“包含”、“包括”等詞,係用以表示存在特定的技術特徵、數值、方法步驟、作業處理、元件以及/或組件,但並不排除可加上更多的技術特徵、數值、方法步驟、作業處理、元件、組件,或以上的任意組合。It must be understood that words such as "comprising" and "including" used in this specification are used to indicate the existence of specific technical features, values, method steps, operations, components and/or components, but do not exclude possible Add more technical characteristics, values, method steps, operation processes, components, components, or any combination of the above.
於權利要求中使用如“第一”、“第二”等詞係用來修飾權利要求中的元件,並非用來表示之間具有優先權順序,先行關係,或者是一個元件先於另一個元件,或者是執行方法步驟時的時間先後順序,僅用來區別具有相同名字的元件。Words such as "first" and "second" used in the claims are used to modify the elements in the claims, and are not used to indicate that there is an order of priority, an antecedent relationship, or that one element is prior to another element , or the chronological order in which the steps of the method are performed, are only used to distinguish elements with the same name.
請一併參閱圖1與圖2,圖1係依據本發明一實施例繪示一種聲紋辨識裝置之方塊圖,圖2係依據本發明另一實施例繪示一種聲紋辨識裝置之方塊圖。Please refer to FIG. 1 and FIG. 2 together. FIG. 1 is a block diagram of a voiceprint recognition device according to an embodiment of the present invention, and FIG. 2 is a block diagram of a voiceprint recognition device according to another embodiment of the present invention. .
於一實施例中,如圖1所示,聲紋辨識裝置100包含一音訊接收單元10、一聲紋辨識單元20、一運算單元30以及一儲存單元40。In one embodiment, as shown in FIG. 1 , the
音訊接收單元10係擷取一音頻資料。聲紋辨識單元20係連接音訊接收單元10以接收上述音頻資料,且可辨識出上述音頻資料內的至少一聲紋特徵。在一些實施例中,上述至少一聲紋特徵可以是特定關鍵詞、聲波波段、聲波波形、聲波頻率等。於一實施例中,聲紋辨識單元20用於辨識音頻資料內的上述至少一聲紋特徵,並且聲紋辨識單元20可包含以下操作中至少一者,例如為濾波、降低雜訊、抑制背景噪音、放大特定聲紋特徵、計算音頻資料之一向量參數、偵測特定關鍵詞、偵測特定聲波波段、偵測特定聲波波形、偵測特定聲波頻率等。The audio receiving
儲存單元40可儲存第一特徵資料44。在一些實施例中,第一特徵資料44可以是符合使用者之任何一相關聲紋資訊,例如為特定關鍵詞、聲波波段、聲波波形、聲波頻率等。在一些實施例中,第一特徵資料44可以是類比訊號、數位訊號、類比/數位混合訊號等資料儲存之模式。The
運算單元30係連接聲紋辨識單元20與儲存單元40,可接收自聲紋辨識單元20的上述至少一聲紋特徵,以及接收自儲存單元40的第一特徵資料44,於運算單元30內判斷上述至少一聲紋特徵與第一特徵資料44是否相符。The
若運算單元30判斷上述至少一聲紋特徵符合第一特徵資料44,則運算單元30將上述至少一聲紋特徵儲存於儲存單元40的一暫存空間42內,且根據暫存空間42內的上述至少一聲紋特徵來訓練第一特徵資料44,以產生一調整後第一特徵資料44,並根據調整後第一特徵資料44進行安全性驗證。舉例來說,上述安全性驗證可以在調整後第一特徵資料44建立完當下立即執行,以更進一步達成一快速安全性驗證之效果。If the calculation unit 30 judges that the above-mentioned at least voiceprint features conform to the
在另一範例中,上述安全性驗證可以待後續收到新的一聲紋特徵,再將此新的一聲紋特徵與調整後第一特徵資料44比對,如此同樣可達到安全性驗證之目的,且相較於前一個範例,可降低誤判斷為「不通過」之機率,提昇整體流程的順暢度,進而提昇使用者體驗。In another example, the above-mentioned security verification can wait for a new voiceprint feature to be received later, and then compare the new voiceprint feature with the adjusted
在一些實施例中,上述判斷的至少一聲紋特徵可對應於使用者之一正常聲紋,或受到年齡、情緒、或身體狀況等因素影響後之異常聲紋,第一特徵資料44則可視為使用者原始預設之聲紋特徵,若至少一聲紋特徵能與第一特徵資料44相符,則使用調整後第一特徵資料44作為聲紋識別標準來進行安全性驗證,且第一特徵資料44可被持續訓練,以持續更新第一特徵資料44可接受之聲紋範疇。In some embodiments, at least the voiceprint characteristics determined above may correspond to a normal voiceprint of the user, or an abnormal voiceprint affected by factors such as age, emotion, or physical condition, and the first
於一實施例中,如圖2所示,聲紋辨識裝置200之運算單元30可進一步包含一驗證計算器32,驗證計算器32之計數值於一初始狀態為0單位,若運算單元30判斷上述至少一聲紋特徵不符第一特徵資料44,則驗證計算器32之計數值增加1單位。In one embodiment, as shown in FIG. 2, the
若上述計算值小於一預定數值,若運算單元30判斷上述至少一聲紋特徵不符第一特徵資料44,則運算單元30將傳送一驗證請求給使用者,且等待上述驗證請求被使用者完成確認後,運算單元30將上述至少一聲紋特徵儲存於暫存空間42內,但儲存單元40不根據上述至少一聲紋特徵來變更第一特徵資料44,而是仍以第一特徵資料44進行安全性驗證。在一些實施例中,上述至少一聲紋特徵可對應於使用者受到年齡、情緒、或身體狀況等因素影響後之異常聲紋,第一特徵資料44則可視為使用者原始預設之聲紋特徵。若異常聲紋的上述至少一聲紋特徵與第一特徵資料44不符,則本發明仍保留第一特徵資料44,並且僅將異常聲紋另外儲存於暫存空間42內,以此可作為一辨識緩衝機制,上述辨識緩衝機制主要避免使用者之上述至少一聲紋特徵僅為聲紋暫時性異常、或聲紋辨識裝置100判斷失誤之可能。If the above-mentioned calculated value is less than a predetermined value, if the
若上述計算值達該預定數值,則運算單元30根據暫存空間42內的上述至少一聲紋特徵與第一特徵資料44產生一第二特徵資料46。在一些實施例中,第二特徵資料46可以是類比訊號、數位訊號、類比/數位混合訊號等資料儲存之模式。If the calculated value reaches the predetermined value, the
於第二特徵資料46產生後,驗證計算器32調整計算值為0單位,運算單元30將第一特徵資料44、第二特徵資料46儲存於儲存單元40,並修改為根據第一特徵資料44、第二特徵資料46,使運算單元30判斷後續接收之至少一聲紋特徵與第一特徵資料44、第二特徵資料46是否相符,以進行安全性驗證。在一些實施例中,上述至少一聲紋特徵可對應於使用者受到年齡、情緒、或身體狀況等因素影響後之異常聲紋,第一特徵資料44則可視為使用者原始預設之聲紋特徵。若異常聲紋的上述至少一聲紋特徵已達一定預定次數(預定時間)而不能符合第一特徵資料44,則本發明仍保留第一特徵資料44,並且另根據暫存空間42內儲存的上述至少一聲紋特徵與第一特徵資料44產生一第二特徵資料46,以此可作為另一辨識緩衝機制,上述辨識緩衝機制主要增加使用者受到年齡、情緒、或身體狀況等因素一定時間影響後的聲紋特徵辨識資料。After the second
於一實施例中,運算單元30於判斷至少一聲紋特徵與第一特徵資料44、第二特徵資料46是否相符時,若上述至少一聲紋特徵符合第二特徵資料46,則運算單元30傳送上述至少一聲紋特徵儲存於暫存空間42內,且根據暫存空間42內的上述至少一聲紋特徵訓練第二特徵資料46,以產生一調整後第二特徵資料46,並根據該調整後第二特徵資料46進行安全性驗證。舉例來說,上述安全性驗證可以在調整後第二特徵資料46建立完當下立即執行,以更進一步達成一快速安全性驗證之效果。In one embodiment, when the
在另一範例中,上述安全性驗證可以待後續收到新的一聲紋特徵,再將此新的一聲紋特徵與調整後第二特徵資料46比對,如此同樣可達到安全性驗證之目的,且相較於前一個範例,可降低誤判斷為「不通過」之機率,提昇整體流程的順暢度,進而提昇使用者體驗。In another example, the above-mentioned security verification can wait for a new voiceprint feature to be received later, and then compare the new voiceprint feature with the adjusted
在一些實施例中,上述至少一聲紋特徵可對應於使用者受到年齡、情緒、或身體狀況等因素影響後之異常聲紋,第二特徵資料46則可視為使用者受一定時間異常影響後的聲紋特徵辨識資料。若異常聲紋的上述至少一聲紋特徵已達一定預定次數(預定時間)而卻與第二特徵資料46相符,則本發明仍保留第一特徵資料44,並且僅判斷為使用者已達一聲紋異常變化後的穩定狀況,而使用調整後第二特徵資料46作為此穩定狀況的聲紋識別標準來進行安全性驗證,且第二特徵資料46可被持續訓練,以持續更新第二特徵資料46可接受之聲紋範疇。In some embodiments, at least the above-mentioned voiceprint characteristics may correspond to the abnormal voiceprint of the user affected by factors such as age, emotion, or physical condition. Voiceprint feature recognition data. If the above-mentioned at least voiceprint characteristics of the abnormal voiceprint have reached a certain predetermined number of times (predetermined time) but match the second
若運算單元30判斷上述至少一聲紋特徵不符第二特徵資料46、但符合第一特徵資料44,則運算單元30自儲存單元40中移除第二特徵資料46。第二特徵資料46移除後,重新根據第一特徵資料44進行安全性驗證。在一些實施例中,上述至少一聲紋特徵可對應於使用者受到年齡、情緒、或身體狀況等因素影響後之異常聲紋特徵,第一特徵資料44可視為使用者原始預設之聲紋特徵,第二特徵資料46則可視為使用者受一定時間異常影響後的聲紋特徵辨識資料。若使用者由造成異常聲紋特徵之狀況恢復時,則可判斷為第二特徵資料46已不適用為一聲紋識別標準,故予以移除第二特徵資料46、恢復為根據第一特徵資料44作為聲紋識別之標準來進行安全性驗證。If the
若運算單元30判斷上述至少一聲紋特徵不符第二特徵資料46且不符第一特徵資料44,則驗證計算器32對上述計數值增加1單位。若上述計算值小於一預定數值,則運算單元30傳送一驗證請求給使用者,且等待上述驗證請求被使用者完成確認後,運算單元30將上述至少一聲紋特徵儲存於暫存空間42內,但儲存單元40不根據上述至少一聲紋特徵變更第一特徵資料44、第二特徵資料46,以根據該第一特徵資料、該第二特徵進行安全性驗證。If the
於一實施例中,若上述計算值達該預定數值,則運算單元30根據暫存空間42內的上述至少一聲紋特徵訓練第二特徵資料46,以產生一調整後第二特徵資料46。產生調整後第二特徵資料46後,驗證計算器32調整該計算值為0單位,且運算單元30判斷至少一聲紋特徵與第一特徵資料44、調整後第二特徵資料46是否相符,以進行安全性驗證。舉例來說,上述安全性驗證可以在調整後第二特徵資料46建立完當下立即執行,以更進一步達成一快速安全性驗證之效果。In one embodiment, if the calculated value reaches the predetermined value, the
在另一範例中,上述安全性驗證可以待後續收到新的一聲紋特徵,再將此新的一聲紋特徵與第一特徵資料44、調整後第二特徵資料46比對,如此同樣可達到安全性驗證之目的,且相較於前一個範例,可降低誤判斷為「不通過」之機率,提昇整體流程的順暢度,進而提昇使用者體驗。In another example, the above-mentioned security verification can wait for a new voiceprint feature to be received later, and then compare the new voiceprint feature with the
在一些實施例中,上述至少一聲紋特徵可對應於使用者受到年齡、情緒、或身體狀況等因素影響後之異常聲紋特徵,且造成異常聲紋特徵之狀況恐已難以恢復。若異常聲紋的上述至少一聲紋特徵已達一定預定次數(預定時間)而不能符合第一特徵資料44、第二特徵資料46,則可判斷為第二特徵資料46已不適用為一聲紋異常變化達穩定狀況的聲紋識別標準,故利用暫存空間42內的聲紋特徵重新訓練第二特徵資料46,產生一調整後第二特徵資料46,使其重新作為聲紋識別之標準來進行安全性驗證。In some embodiments, at least the above-mentioned voiceprint characteristics may correspond to abnormal voiceprint characteristics of the user affected by factors such as age, emotion, or physical condition, and the conditions causing abnormal voiceprint characteristics may be difficult to recover. If the above-mentioned at least voiceprint characteristics of the abnormal voiceprint have reached a certain predetermined number of times (predetermined time) and cannot meet the first
於一實施例中,聲紋辨識裝置100、200可設置於透過網路執行金融交易的個人電腦、膝上型電腦、蜂窩電話、相機電話、智慧型電話、個人數位助理、媒體播放機、導航設備、電子郵件收發設備、遊戲控制台、平板電腦、可穿戴設備或者這些設備中的任意幾種設備的組合。In one embodiment, the
於一實施例中,運算單元30可以是一中央處理器(central processing unit,CPU),亦可配置為其他運算能力足夠的元件。In an embodiment, the
於一實施例中,儲存單元40可以是一伺服器(server),用於儲存一或多個公司、銀行或機構的聲紋辨識資料庫。於另一實施例中,儲存單元40可以是一任何型態的固定式或可移動式的隨機存取記憶體(random access memory,RAM)、唯讀記憶體(read-only memory,ROM)、快閃記憶體(flash memory)、硬碟(hard disk drive,HDD)、固態硬碟(solid state drive,SSD)或類似元件或上述元件的組合。In one embodiment, the
於一實施例中,上述驗證請求所包含的使用者身份特徵之識別方法可以是簡訊、傳真、電子郵件、應用程式、數位簽章、或基於其他生物學特徵(例如為臉部影像、指紋、虹膜等辨識)等。In one embodiment, the identification method of the user's identity included in the verification request can be text message, fax, email, application program, digital signature, or based on other biological characteristics (such as facial image, fingerprint, iris, etc. identification), etc.
於一實施例中,儲存單元40可包含一深度學習模型,用以訓練第一特徵資料44、第二特徵資料46,上述深度學習模型可包括各種類型的神經網路模型中的至少一者,所述各種類型的神經網路模型包括卷積神經網路(convolution neural network,CNN)、區域卷積神經網路(region with convolution neural network,R-CNN)、區域建議網絡(region proposal network,RPN)、遞迴神經網路(recurrent neural network,RNN)、基於堆疊的深度神經網路(stacking-based deep neural network,S-DNN)、狀態空間動態神經網路(state-space dynamic neural network,S-SDNN)、反卷積網路(deconvolution network)、深度置信網路(deep belief network,DBN)、受限玻爾茲曼機(restricted Boltzmann machine,RBM)、全卷積網路(fully convolutional network)、長短期記憶體(long short-term memory,LSTM)網路以及分類網路(classification network)。In one embodiment, the
請一併參閱圖3與圖4,圖3係根據本發明之一實施例繪示一種聲紋辨識方法之流程圖,圖4係根據本發明之另一實施例繪示一種聲紋辨識方法之流程圖。Please refer to FIG. 3 and FIG. 4 together. FIG. 3 is a flow chart of a voiceprint recognition method according to an embodiment of the present invention, and FIG. 4 is a flowchart of a voiceprint recognition method according to another embodiment of the present invention. flow chart.
於一實施例中,如圖3所示,於步驟S1中,藉由音訊接收單元10擷取一音頻資料。In one embodiment, as shown in FIG. 3 , in step S1 , an audio data is captured by the
於步驟S2中,聲紋辨識單元20接收音訊接收單元10所傳送的上述音頻資料,且辨識出上述音頻資料內的至少一聲紋特徵。In step S2, the
於步驟S3中,運算單元30接收聲紋辨識單元20所傳送的上述至少一聲紋特徵、接收儲存單元40所傳送的第一特徵資料44,用以比對上述聲紋特徵與第一特徵資料44是否相符。In step S3, the
於步驟S4中,若上述至少一聲紋特徵符合第一特徵資料44,則運算單元30將上述至少一聲紋特徵儲存於暫存空間42內,且運算單元30根據暫存空間42內的上述至少一聲紋特徵訓練第一特徵資料44,以產生一調整後第一特徵資料44,以根據調整後第一特徵資料44進行安全性驗證。舉例來說,上述安全性驗證可以在調整後第一特徵資料44建立完當下立即執行,以更進一步達成一快速安全性驗證之效果。In step S4, if the above-mentioned at least voiceprint feature matches the
在另一範例中,上述安全性驗證可以待後續收到新的一聲紋特徵,再將此新的一聲紋特徵與調整後第一特徵資料44比對,如此同樣可達到安全性驗證之目的,且相較於前一個範例,可降低誤判斷為「不通過」之機率,提昇整體流程的順暢度,進而提昇使用者體驗。In another example, the above-mentioned security verification can wait for a new voiceprint feature to be received later, and then compare the new voiceprint feature with the adjusted
於步驟S5中,運算單元30可包含一驗證計算器32,驗證計算器32之計數值於一初始狀態為0單位,若上述聲紋特徵不符第一特徵資料44,則增加驗證計算器32之計數值1單位。若上述計算值小於一預定數值,則運算單元30將傳送一驗證請求給使用者,當上述驗證請求被使用者完成確認後,則運算單元30將上述至少一聲紋特徵儲存於暫存空間42內,但不根據上述至少一聲紋特徵變更第一特徵資料44,以根據第一特徵資料44進行安全性驗證。In step S5, the
於步驟S6中,若上述至少一聲紋特徵不符第一特徵資料44,且上述計算值達一預定數值,則運算單元30使暫存空間42內的至少一聲紋特徵與第一特徵資料44產生一第二特徵資料46。In step S6, if the above-mentioned at least voiceprint feature does not match the
於步驟S7中,第二特徵資料46產生後,驗證計算器32調整計算值為0單位,且運算單元30修正成根據第一特徵資料44、第二特徵資料46來比對至少一聲紋特徵與第一特徵資料44、第二特徵資料46是否相符,以進行安全性驗證。In step S7, after the
於一實施例中,如圖4所示,於步驟S8中,運算單元30修正成根據第一特徵資料44、第二特徵資料46來比對至少一聲紋特徵與第一特徵資料44、第二特徵資料46是否相符,而其中運算單元30先根據第二特徵資料46來比對至少一聲紋特徵與第二特徵資料46是否相符。In one embodiment, as shown in FIG. 4 , in step S8, the
於步驟S9中,若上述聲紋特徵符合第二特徵資料46,則運算單元30將上述至少一聲紋特徵儲存於暫存空間42內,且儲存單元40根據暫存空間42內的上述至少一聲紋特徵訓練第二特徵資料46,以產生一調整後第二特徵資料46,並根據調整後第二特徵資料46進行安全性驗證。舉例來說,上述安全性驗證可以在調整後第二特徵資料46建立完當下立即執行,以更進一步達成一快速安全性驗證之效果。In step S9, if the above-mentioned voiceprint feature matches the
在另一範例中,上述安全性驗證可以待後續收到新的一聲紋特徵,再將此新的一聲紋特徵與調整後第二特徵資料46比對,如此同樣可達到安全性驗證之目的,且相較於前一個範例,可降低誤判斷為「不通過」之機率,提昇整體流程的順暢度,進而提昇使用者體驗。In another example, the above-mentioned security verification can wait for a new voiceprint feature to be received later, and then compare the new voiceprint feature with the adjusted
於步驟S10中,若上述聲紋特徵不符第二特徵資料46,則運算單元30再根據第一特徵資料44來比對上述聲紋特徵與第一特徵資料44是否相符。In step S10 , if the above-mentioned voiceprint features do not match the
於步驟S11中,若上述聲紋特徵不符第二特徵資料46、但符合第一特徵資料44,則運算單元30移除第二特徵資料46,以根據第一特徵資料44進行安全性驗證。In step S11 , if the above-mentioned voiceprint features do not match the
於步驟S12中,驗證計算器32之計數值於一初始狀態為0單位,若上述聲紋特徵不符第二特徵資料46且不符第一特徵資料44,則增加驗證計算器32之計數值1單位。若上述計算值小於一預定數值,則傳送一驗證請求給使用者,上述驗證請求被使用者完成確認後,運算單元30傳送上述聲紋特徵至暫存空間42內儲存,但儲存單元40不根據上述至少一聲紋特徵變更第一特徵資料44、第二特徵資料46,以根據第一特徵資料44、第二特徵資料46進行安全性驗證。In step S12, the count value of the
於步驟S13中,若上述至少一聲紋特徵不符第二特徵資料46且不符第一特徵資料44,且上述計算值達一預定數值,則運算單元30根據暫存空間42的上述至少一聲紋特徵訓練第二特徵資料46,以產生一調整後第二特徵資料46。In step S13, if the above-mentioned at least voiceprint feature does not match the
於步驟S14中,產生調整後第二特徵資料46後,該驗證計算器調整該計算值為0單位,運算單元30修正成根據第一特徵資料44、調整後第二特徵資料46來判斷至少一聲紋特徵與第一特徵資料44、調整後第二特徵資料46是否相符,以進行安全性驗證。舉例來說,上述安全性驗證可以在調整後第二特徵資料46建立完當下立即執行,以更進一步達成一快速安全性驗證之效果。In step S14, after the adjusted
在另一範例中,上述安全性驗證可以待後續收到新的一聲紋特徵,再將此新的一聲紋特徵與第一特徵資料44、調整後第二特徵資料46比對,如此同樣可達到安全性驗證之目的,且相較於前一個範例,可降低誤判斷為「不通過」之機率,提昇整體流程的順暢度,進而提昇使用者體驗。In another example, the above-mentioned security verification can wait for a new voiceprint feature to be received later, and then compare the new voiceprint feature with the
於一實施例中,聲紋辨識方法300、400可設置於電子金融交易系統,例如為自動櫃員機(ATM)、網路銀行、或其他金融交易系統上。In one embodiment, the voiceprint recognition methods 300 and 400 can be set in an electronic financial transaction system, such as an automatic teller machine (ATM), online banking, or other financial transaction systems.
於一實施例中,上述驗證請求所包含的使用者身份特徵之識別方法可以是簡訊、傳真、電子郵件、應用程式、數位簽章、或基於其他生物學特徵(例如為臉部影像、指紋、虹膜等辨識)等。In one embodiment, the identification method of the user's identity included in the verification request can be text message, fax, email, application program, digital signature, or based on other biological characteristics (such as facial image, fingerprint, iris, etc. identification), etc.
於一實施例中,聲紋辨識方法300、400中之聲紋辨識裝置可包含一深度學習模型,用以訓練第一特徵資料44、第二特徵資料46,上述深度學習模型可包括各種類型的神經網路模型中的至少一者,所述各種類型的神經網路模型包括卷積神經網路(convolution neural network,CNN)、區域卷積神經網路(region with convolution neural network,R-CNN)、區域建議網絡(region proposal network,RPN)、遞迴神經網路(recurrent neural network,RNN)、基於堆疊的深度神經網路(stacking-based deep neural network,S-DNN)、狀態空間動態神經網路(state-space dynamic neural network,S-SDNN)、反卷積網路(deconvolution network)、深度置信網路(deep belief network,DBN)、受限玻爾茲曼機(restricted Boltzmann machine,RBM)、全卷積網路(fully convolutional network)、長短期記憶體(long short-term memory,LSTM)網路以及分類網路(classification network)。In one embodiment, the voiceprint recognition device in the voiceprint recognition methods 300 and 400 may include a deep learning model for training the
本發明對於電子裝置上的金融交易系統可做為一道輔助認證的功能。使用者於登入系統環境欲進行金融交易等動作時,藉由讀取裝置使用時所記錄的聲紋圖譜,透過深度學習、加入遞歸神經網路來訓練聲紋圖譜模型,藉此改善認證的品質,讓模型可以更準確地辨認使用者本人的指令,進而將此項認證應用於各種金融交易系統中,例如為行動銀行(網路銀行)等。The invention can be used as an auxiliary authentication function for the financial transaction system on the electronic device. When the user logs into the system environment and intends to perform financial transactions and other actions, by reading the voiceprint recorded during the use of the device, the voiceprint model is trained through deep learning and adding a recurrent neural network, thereby improving the quality of authentication , so that the model can more accurately identify the user's own instructions, and then apply this authentication to various financial transaction systems, such as mobile banking (online banking), etc.
惟以上所述者,僅為本發明之實施例而已,當不能以此限定本發明實施之範圍,凡是依本發明申請專利範圍及專利說明書內容所作之簡單的等效變化與修飾,皆仍屬本發明專利涵蓋之範圍內。But what is described above is only an embodiment of the present invention, and should not limit the scope of the present invention. All simple equivalent changes and modifications made according to the patent scope of the present invention and the content of the patent specification are still within the scope of the present invention. Within the scope covered by the patent of the present invention.
100,200:聲紋辨識裝置 300,400:聲紋辨識方法 10:音訊接收單元 20:聲紋辨識單元 30:運算單元 32:驗證計算器 40:儲存單元 42:暫存空間 44:第一特徵資料 46:第二特徵資料 S1~S14:步驟 S1’,S2’:步驟 100,200: voiceprint recognition device 300,400: voiceprint recognition method 10: Audio receiving unit 20:Voiceprint recognition unit 30: Operation unit 32: Verification Calculator 40: storage unit 42: temporary storage space 44: The first feature data 46: Second characteristic data S1~S14: Steps S1', S2': step
圖1係依據本發明一實施例繪示一種聲紋辨識裝置之方塊圖。FIG. 1 is a block diagram of a voiceprint recognition device according to an embodiment of the present invention.
圖2係依據本發明另一實施例繪示一種聲紋辨識裝置之方塊圖。FIG. 2 is a block diagram of a voiceprint recognition device according to another embodiment of the present invention.
圖3係依據本發明之一實施例繪示一種聲紋辨識方法之流程圖。FIG. 3 is a flowchart illustrating a voiceprint recognition method according to an embodiment of the present invention.
圖4係依據本發明之另一實施例繪示一種聲紋辨識方法之流程圖。FIG. 4 is a flowchart illustrating a voiceprint recognition method according to another embodiment of the present invention.
無none
400:聲紋辨識方法 400:Voiceprint recognition method
S1~S14:步驟 S1~S14: Steps
S1’,S2’:步驟 S1', S2': step
Claims (6)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW112103776A TWI817897B (en) | 2021-09-08 | 2021-09-08 | Low-noise voiceprint identification device for financial transaction system and method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW112103776A TWI817897B (en) | 2021-09-08 | 2021-09-08 | Low-noise voiceprint identification device for financial transaction system and method thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
TW202328951A true TW202328951A (en) | 2023-07-16 |
TWI817897B TWI817897B (en) | 2023-10-01 |
Family
ID=88147483
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW112103776A TWI817897B (en) | 2021-09-08 | 2021-09-08 | Low-noise voiceprint identification device for financial transaction system and method thereof |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI817897B (en) |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103971690A (en) * | 2013-01-28 | 2014-08-06 | 腾讯科技(深圳)有限公司 | Voiceprint recognition method and device |
TWI639114B (en) * | 2017-08-30 | 2018-10-21 | 元鼎音訊股份有限公司 | Electronic device with a function of smart voice service and method of adjusting output sound |
CN108040032A (en) * | 2017-11-02 | 2018-05-15 | 阿里巴巴集团控股有限公司 | A kind of voiceprint authentication method, account register method and device |
CN108269575B (en) * | 2018-01-12 | 2021-11-02 | 平安科技(深圳)有限公司 | Voice recognition method for updating voiceprint data, terminal device and storage medium |
-
2021
- 2021-09-08 TW TW112103776A patent/TWI817897B/en active
Also Published As
Publication number | Publication date |
---|---|
TWI817897B (en) | 2023-10-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10262659B2 (en) | Hotword recognition | |
KR101917791B1 (en) | Hotword recognition | |
US9934783B2 (en) | Hotword recognition | |
US10672403B2 (en) | Age compensation in biometric systems using time-interval, gender and age | |
US9646613B2 (en) | Methods and systems for splitting a digital signal | |
US11252152B2 (en) | Voiceprint security with messaging services | |
CN105723450A (en) | Envelope comparison for utterance detection | |
WO2019062721A1 (en) | Training method for voice identity feature extractor and classifier and related devices | |
US20230386506A1 (en) | Self-supervised speech representations for fake audio detection | |
US20200082830A1 (en) | Speaker recognition | |
TWM622203U (en) | Voiceprint identification device for financial transaction system | |
US10446138B2 (en) | System and method for assessing audio files for transcription services | |
US11841932B2 (en) | System and method for updating biometric evaluation systems | |
KR20190142056A (en) | Voice recognition otp authentication method using machine learning and system thereof | |
TWI787996B (en) | Voiceprint identification device for financial transaction system and method thereof | |
TWI817897B (en) | Low-noise voiceprint identification device for financial transaction system and method thereof | |
Kuznetsov et al. | Methods of countering speech synthesis attacks on voice biometric systems in banking | |
US11582336B1 (en) | System and method for gender based authentication of a caller | |
US11227610B1 (en) | Computer-based systems for administering patterned passphrases | |
Evsyukov et al. | Methods of protection in speaker verification systems | |
US20230131359A1 (en) | System and Method for Generating Synthetic Cohorts Using Generative Modeling | |
Nakajima et al. | Speaker Verification Based on Synchronous Speech and Video Features | |
US20230386453A1 (en) | Method for detecting an audio adversarial attack with respect to a voice command processed byan automatic speech recognition system, corresponding device, computer program product and computer-readable carrier medium | |
WO2023158972A1 (en) | Speaker verification with multitask speech models | |
CN115641853A (en) | Voiceprint recognition method and device, computer readable storage medium and terminal |