Speech recognition device and method

Abstract

The invention provides a voice recognition device and method. The voice recognition device comprises an attack defense module and a speaker verification module. The attack defense module has a first deep learning model. The first deep learning model determines whether the first input speech is true speech or false speech. When the first input voice is real voice, the attack defense module outputs a second input voice. The speaker verification module has a second deep learning model. The second deep learning model processes a second input speech to generate a speech feature. The speaker verification module determines whether the first input voice is from a legal person according to the voice feature. When the first input voice comes from a legal person, the speaker verification module performs a specific action. When the first input voice is not from a legal person, the speaker verification module does not perform a specific action.

Description

Technical Field

[0001] The present invention relates to a speech recognition device, and in particular to a speech recognition device with attack and defense capabilities. Background Art

[0002] For example, in order to achieve a higher level of security, a smart access control system that focuses on security may require multiple systems (such as facial, fingerprint, and voice recognition) to enter the room. However, such a system consumes huge power and increases the cost a lot.

[0003] Furthermore, with the rapid development of science and technology, some malicious attack methods have also made greater progress. For example, voice attacks include speech synthesis, pre-recorded playback, and even fake audio generation. If the attack methods are not updated frequently, they are exposed to risks at all times. Summary of the invention

[0004] One embodiment of the present invention provides a speech recognition device, including an attack defense module and a speaker verification module. The attack defense module is used to filter out a false voice, and includes a processing circuit and an inference circuit. The processing circuit processes a first input voice to generate an audio signal. The inference circuit receives the audio signal and processes the audio signal using a first deep learning model to determine whether the first input voice is a true voice or a false voice. When the first input voice is a true voice, the inference circuit outputs a second input voice. The speaker verification module includes a feature extraction circuit, a comparison circuit and a decision circuit. The feature extraction circuit uses a second deep learning model to process the second input voice to generate a voice feature. The comparison circuit compares the voice feature with at least one preset feature to generate a comparison result. The decision circuit determines whether the comparison result is greater than a threshold. When the comparison result is greater than the threshold, the decision circuit performs a specific action. When the comparison result is not greater than the threshold, the decision circuit does not perform a specific action.

[0005] The present invention further provides a speech recognition method, including receiving and processing an input speech to generate an audio signal; using a first deep learning model to process the audio signal to determine whether the input speech is a true speech or a false speech; when the input speech is a true speech, using a second deep learning model to process the input speech to generate a speech feature; comparing the speech feature with at least one preset feature to generate a comparison result; determining whether the comparison result is greater than a threshold; when the comparison result is greater than the threshold, performing a specific action; and when the comparison result is not greater than the threshold, not performing the specific action.

[0006] The speech recognition method of the present invention can be implemented by the speech recognition device of the present invention, which is hardware or firmware that can perform specific functions, or can be recorded in a recording medium in the form of program code and implemented in combination with specific hardware. When the program code is loaded and executed by an electronic device, processor, computer or machine, the electronic device, processor, computer or machine becomes an attack defense module and a speaker verification module for implementing the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 FIG. 4 is a schematic diagram of a speech recognition device according to the present invention.

[0008] Figure 2 Schematic diagram of the attack defense module of the present invention.

[0009] Figure 3A Schematic diagram of a speaker verification module of the present invention.

[0010] Figure 3B FIG. 4 is another schematic diagram of the speaker verification module of the present invention.

[0011] Figure 4 It is a schematic diagram of the attack defense operation of the present invention.

[0012] Figure 5 Schematic diagram of the voice verification operation of the present invention.

[0013] Figure 6 It is a flowchart of the speech recognition method of the present invention.

[0014] Reference numerals:

[0015] 100: Voice recognition device

[0016] 110: Attack Defense Module

[0017] 120: Speaker Verification Module

[0018] 130: Receiving circuit

[0019] SP_IN1~SP_IN3: Input voice

[0020] SP_RL: True Voice

[0021] SP_SF: Fake voice

[0022] ST: trigger signal

[0023] 210: Processing circuit

[0024] 220: Speculation Circuit

[0025] ADS_1~ADS_3: audio signal

[0026] 230, 350: Input and output interface

[0027] DA_1, DA_2: Update data

[0028] 221, 311: Deep learning models

[0029] 310: Feature Extraction Circuit

[0030] 330: Comparison circuit

[0031] 340: Decision circuit

[0032] SP_EM: Speech Features

[0033] S_CM: comparison results

[0034] SP_PR: Preset Features

[0035] S611~S619: Steps DETAILED DESCRIPTION

[0036] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the following examples are specifically cited and described in detail with the accompanying drawings. The present invention specification provides different examples to illustrate the technical features of different implementations of the present invention. Among them, the configuration of each element in the example is for illustrative purposes and is not intended to limit the present invention. In addition, some of the figure numbers in the examples are repeated for the purpose of simplifying the description and do not mean the correlation between different examples.

[0037] Figure 1Schematic diagram of the speech recognition device of the present invention. As shown in the figure, the speech recognition device 100 includes an attack defense module 110 and a speaker verification module 120. In a possible embodiment, the speech recognition device 100 is a door access control speech recognition system. In this example, the speech recognition device 100 first determines whether an input speech SP_IN1 is a real speech. If the input speech SP_IN1 is not a real speech, for example, the input speech SP_IN1 is a spoof speech, the speech recognition device 100 ignores the input speech SP_IN1. If the input speech SP_IN1 is a real speech, the speech recognition device 100 then determines whether the input speech SP_IN1 conforms to the registered preset speech. If so, it means that the person who issued the input speech SP_IN1 is a legal person. Therefore, the speech recognition device 100 performs a specific action according to the input speech SP_IN1, such as opening the door. If the input speech SP_IN1 does not conform to the registered preset speech, it means that the person who issued the input speech SP_IN1 is an illegal person. Therefore, the speech recognition device 100 does not perform a specific action. In some embodiments, although the speech recognition device 100 does not perform a specific action, the speech recognition device 100 performs an alert action, such as issuing an alert sound to notify relevant personnel that someone is trying to enter the room.

[0038] In one possible embodiment, the speech recognition device 100 further includes a receiving circuit 130. The receiving circuit 130 receives an external voice. The external voice may be a real voice SP_RL or a fake voice SP_SF. Regardless of whether the external voice is a real voice SP_RL or a fake voice SP_SF, the receiving circuit 130 uses the external voice as an input voice SP_IN1 and provides the input voice SP_IN1 to the attack defense module 110. In some embodiments, the receiving circuit 130 has a microphone. The present invention does not limit the architecture of the receiving circuit 130. In one possible embodiment, the receiving circuit 130 is a recorder.

[0039] The attack defense module 110 receives the input voice SP_IN1 and determines the type of the input voice SP_IN1. When the input voice SP_IN1 is a real voice, the attack defense module 110 provides an input voice SP_IN2 according to the input voice SP_IN1. In some embodiments, the attack defense module 110 directly uses the input voice SP_IN1 as the input voice SP_IN2. When the input voice SP_IN1 is a fake voice, the attack defense module 110 stops providing the input voice SP_IN2.

[0040] In one possible embodiment, the true voice is the voice emitted by a living body (such as a user), and the false voice is the voice generated by a voice generator. The voice generator may generate a false voice using a voice synthesis method, a pre-recorded playback method, or a false audio generation method. A person with intentions may approach the receiving circuit 130 and then play the false voice. In this example, although the false voice is also emitted by a living body, it is not the voice spoken directly by the living body to the receiving circuit 130. Therefore, as long as the voice is not spoken by the living body to the receiving circuit 130, it is called a false voice. In other embodiments, the voice generator may capture part of the voice of the living body, rearrange it, and even use the upper and lower limits of human hearing to simulate an attack voice (i.e., a false voice). Therefore, as long as the voice is played, it is called a false voice.

[0041] In the present embodiment, the attack defense module 110 has a first deep learning model. The parameters of the first deep learning model are provided by an external device. The external device performs training of real speech and false speech, and then writes the trained parameters into the first deep learning model of the attack defense module 110. Therefore, the attack defense module 110 has the ability to distinguish whether the input speech SP_IN1 is a false speech or a real speech. In this example, the external device may update the parameters of the first deep learning model regularly or from time to time, so that the speech recognition device 100 is more robust and effective in resisting attacks by new false speech. When a new type of attack occurs, the external device performs training according to the attack data, and then uses the trained parameters to update the parameters of the first deep learning model. The first deep learning model can distinguish new attacking speech.

[0042] In some embodiments, the attack defense module 110 performs machine learning to distinguish advanced speech synthesis. The attack defense module 110 may train the attack speech generated by the fake audio generator and the real human voice, and then write the parameters generated by the training into the first deep learning model, so that the real voice and the hypothetical voice can be distinguished, and malicious attack speech can be blocked.

[0043] The speaker verification module 120 receives the input voice SP_IN2 and determines whether the input voice SP_IN2 is the voice of a registered user. If the input voice SP_IN2 is the voice of a registered user, the speaker verification module 120 performs a specific action. In one possible embodiment, the speaker verification module 120 enables a trigger signal ST to open the door lock. If the input voice SP_IN2 is not the voice of a registered user, the speaker verification module 120 does not perform a specific action. For example, the speaker verification module 120 may not enable the trigger signal ST. Therefore, the door lock continues to be locked. In other embodiments, the speaker verification module 120 may emit a warning sound to notify a specific person that someone is trying to enter the house.

[0044] In some embodiments, during a registration period, the speaker verification module 120 receives an input voice SP_IN3. The input voice SP_IN3 is the voice of a legitimate person. The speaker verification module 120 captures the voice-point of the input voice SP_IN3 and uses the voice-point of the input voice SP_IN3 as a preset feature. In this example, the speaker verification module 120 compares the voice-points of the input voices SP_IN2 and SP_IN3. When the voice-point of the input voice SP_IN2 is similar to the preset feature, it indicates that the input voice SP_IN1 is emitted by a legitimate person. Therefore, the speaker verification module 120 performs a specific action, such as opening a door lock.

[0045] In one possible embodiment, the speaker verification module 120 has a second deep learning model. The parameters of the second deep learning model are provided by an external device. The external device uses multiple speech as training data to learn how to extract the speaker's voice features from the speech. The external device writes the trained parameters into the second deep learning model of the speaker verification module 120. Therefore, the second deep learning model has the ability to extract voice features. During a registration period, the second deep learning model extracts the voice features of at least one legitimate person and records the extracted voice features in a database. In some embodiments, when the second deep learning model extracts the voice features of multiple legitimate persons, the database records multiple voice features. Each voice feature corresponds to the voice of a legitimate person.

[0046] Figure 2Schematic diagram of the attack defense module 110 of the present invention. The attack defense module 110 includes a processing circuit 210 and an inference circuit 220. The processing circuit 210 processes the input voice SP_IN1 to generate an audio signal ADS_1. The present invention does not limit the architecture of the processing circuit 210. In one possible embodiment, the processing circuit 210 has the ability to extract features. For example, the processing circuit 210 may extract the constant Q cepstral coefficient (CQCC) or the Mel cepstral coefficient (MFCC) of the input voice SP_IN1 to generate an extraction result. In this example, the processing circuit 210 uses the extraction result as the audio signal ADS_1.

[0047] The inference circuit 220 receives the audio signal ADS_1 and processes the audio signal ADS_1 using a deep learning model 221 to determine whether the input voice SP_IN1 is a real voice or a fake voice. When the input voice SP_IN1 is a real voice, the inference circuit 220 outputs an input voice SP_IN2. In some embodiments, the inference circuit 220 uses the input voice SP_IN1 or the audio signal ADS_1 as the input voice SP_IN2. When the input voice SP_IN1 is a fake voice, the inference circuit 220 does not provide the input voice SP_IN2. In one possible embodiment, the inference circuit 220 has a storage device for storing the deep learning model 211.

[0048] In some embodiments, the attack defense module 110 further includes an input-output interface 230. The input-output interface 230 is used to receive an update data DA_1. The update data DA_1 is provided by an external device. The external device trains a training data. The training data includes at least one real voice and at least one fake voice. After training, the external device generates at least one parameter and uses the parameter as the update data DA_1. During an update period, when the external device is coupled to the input-output interface 230, the inference circuit 220 receives the update data DA_1 through the input-output interface 230. The inference circuit 220 writes the update data DA_1 into the deep learning model 221 to update at least one parameter of the deep learning model 221.

[0049] The present invention does not limit the type of the input / output interface 230. In one possible embodiment, the input / output interface 230 is connected to an external device via a connection line. In another possible embodiment, the input / output interface 230 is a wireless receiver that receives the update data DA_1 from the external device.

[0050] Figure 3ASchematic diagram of the speaker verification module 120 of the present invention. As shown in the figure, the speaker verification module 120 includes a feature extraction circuit 310, a comparison circuit 330 and a decision circuit 340. In the present embodiment, the speaker verification module 120 is used to determine whether the input voice SP_IN2 from the attack defense module 110 is a legal voice. When the input voice SP_IN1 is a legal voice, the speaker verification module 120 performs a specific action, such as starting. When the input voice SP_IN1 is an illegal voice, the speaker verification module 120 does not perform a specific action, or performs a warning action, such as issuing a warning sound to notify the people in the room.

[0051] The feature extraction circuit 310 uses a deep learning model 311 to process the input speech SP_IN2 to generate a speech feature SP_EM. In one possible embodiment, the input speech SP_IN2 is the input speech SP_IN1. In this example, when the attack defense module 110 determines that the input speech SP_IN1 is a true speech, the attack defense module 110 directly uses the audio signal ADS_1 as the input speech SP_IN2. In this embodiment, the deep learning model 311 extracts the speaker's feature (speaker embedding). In one possible embodiment, the feature extraction circuit 310 has a storage device for storing the deep learning model 311.

[0052] The comparison circuit 330 compares the speech feature SP_EM with at least one preset feature SP_PR to generate a comparison result S_CM. In some embodiments, the comparison result S_CM indicates the similarity between the speech feature SP_EM and the preset feature SP_PR.

[0053] The decision circuit 340 determines whether the comparison result S_CM is greater than a threshold. When the comparison result S_CM is greater than the threshold, the decision circuit 340 performs a specific action, such as enabling the trigger signal ST. When the comparison result S_CM is not greater than the threshold, the decision circuit 340 does not perform a specific action, such as not enabling the trigger signal ST.

[0054] In some embodiments, the speaker verification module 120 further includes a storage circuit 320. The storage circuit 320 is used to store the preset feature SP_PR. In one possible embodiment, the preset feature SP_PR is provided by the feature extraction circuit 310. The present invention does not limit the number of preset features SP_PR. The storage circuit 320 may store more preset features SP_PR. In this example, each preset feature SP_PR represents a voice feature of a legitimate person (such as a family member). The present invention does not limit the type of the storage circuit 320. In one possible embodiment, the storage circuit 320 is a non-volatile memory.

[0055] In one possible embodiment, the parameters of the deep learning model 311 are obtained by an external device (not shown). The external device uses multiple voices as training data to learn to extract voice features. The external device writes the trained parameters into the deep learning model 311. Therefore, the deep learning model 311 has the ability to extract voice features.

[0056] In other embodiments, the speaker verification module 120 further includes an input-output interface 350. The input-output interface 350 is used to receive an update data DA_2. When an external device is coupled to the input-output interface 350, the speaker verification module 120 enters an update mode. In the update mode, the feature extraction circuit 310 updates at least one parameter of the deep learning model 311 according to the update data DA_2. The present invention does not limit the type of the input-output interface 350. In one possible embodiment, the input-output interface 350 is connected to the external device via a connecting line. In another possible embodiment, the input-output interface 350 is a wireless receiver that receives the update data DA_2 from the external device.

[0057] Figure 3B FIG. 1 is another schematic diagram of the speaker verification module 120 of the present invention. Figure 3B resemblance Figure 3A , the difference is that, Figure 3B A processing circuit 360 is added. The processing circuit 360 processes the input voice SP_IN2 to generate an audio signal ADS_2. Since the characteristics of the processing circuit 360 are similar to Figure 2 The processing circuit 210 is not described in detail.

[0058] The feature extraction circuit 310 receives the audio signal ADS_2 and processes the audio signal ADS_2 using the deep learning model 311 to generate the speech feature SP_EM. In this embodiment, when the attack defense module 110 knows that the input speech SP_IN1 is a real speech, the attack defense module 110 uses the input speech SP_IN1 as the input speech SP_IN2.

[0059] In some embodiments, during an initial period (or registration period), the processing circuit 360 processes the input voice SP_IN3 to generate an audio signal ADS_3. In this example, the input voice SP_IN3 is the voice of a legitimate person. The deep learning model 311 extracts the voice features of the audio signal ADS_3 and records the extraction results (i.e., the preset features SP_PR) in the storage circuit 320.

[0060] In one possible embodiment, a legitimate person Figure 1The receiving circuit 130 speaks. The receiving circuit 130 provides the voice of the legal person (i.e., the input voice SP_IN3) to the processing circuit 360. In some embodiments, the receiving circuit 130 may receive the voices of five legal persons. Therefore, the storage circuit 320 records the five-stroke preset features SP_PR, which correspond to the five legal persons respectively. In this example, the comparison circuit 320 compares the voice feature SP_EM with the five-stroke preset features SP_PR, and provides five comparison results to the decision circuit 340. The decision circuit 340 compares each comparison result with a threshold.

[0061] Figure 4 This is a schematic diagram of the attack defense operation of the present invention. In a training phase, many different training data are provided. The training data are real speech and fake speech. Real speech is the sound made by a living body. Fake speech includes speech obtained using various methods, such as generating sounds similar to living body speech. In addition, there are recording and playback, capturing part of the speech, and surround and stereo sound. Next, the features of the training data are extracted, and then speech training (such as machine learning) is performed. Through multiple speech trainings, at least one parameter can be generated. The parameter is then written into a first deep learning model.

[0062] In an inference phase, an inference data is received. In one possible embodiment, the inference data is an unknown voice. The unknown voice may come from a recorder. Then, the features of the inference data are extracted to generate a first extraction result. The first extraction result is processed using a first deep learning model to determine whether the inference data is a true voice or a false voice. In one possible embodiment, if the inference data is a true voice, it is verified whether the true voice is uttered by a legitimate person. If the inference data is a false voice, the false voice is not verified.

[0063] Figure 5 Schematic diagram of the voice verification operation of the present invention. In a training phase, a plurality of different training data are provided. In a possible embodiment, each training data is a real voice. Next, the features of the training data are extracted, and then voice training is performed. Through multiple voice trainings, at least one parameter can be generated. The parameter is then written into a second deep learning model.

[0064] In an enrollment phase, at least one enrollment data is received. Then, features of the enrollment data are extracted to generate a second extraction result. The second extraction result is processed using a second deep learning model to obtain speech features (or preset features) of the enrollment data. In a possible embodiment, the speech features of the enrollment data are stored in a database.

[0065] In a verification phase, a verification data is received. In one possible embodiment, the verification data is an unknown real voice, such as Figure 4 The first deep learning model is used to judge the true speech. Then, the features of the verification data are extracted to produce a third extraction result. The third extraction result is processed by the second deep learning model to produce an unknown voice feature. The unknown voice feature is compared with the preset features in the database to determine the similarity between the unknown voice feature and each preset feature. Next, it is determined whether the similarity between the unknown voice feature and each preset feature exceeds a threshold. If so, it means that the verification data is a voice emitted by a legitimate person, so the verification data is accepted and a specific action is performed. If not, it means that the verification data is not a voice emitted by a legitimate person, so the verification data is rejected and the specific action is not performed.

[0066] Figure 6 The figure is a flow chart of the speech recognition method of the present invention. The speech recognition method of the present invention can be implemented by program code. When the program code is loaded and executed by a machine, the machine becomes a speech recognition device for implementing the present invention.

[0067] First, an input speech is received and processed to generate an audio signal (step S611). In one embodiment, step S611 extracts constant Q cepstral coefficients or Mel cepstral coefficients of the input speech, and then uses the extraction result as the audio signal.

[0068] Next, a first deep learning model is used to process the audio signal (step S612). In one possible embodiment, the parameters of the first deep learning model are provided by an external device. The external device performs training of real speech and fake speech, and then writes the trained parameters into the first deep learning model. Therefore, according to the output result of the first deep learning model, the ability of the input speech to be a fake speech or a real speech can be known.

[0069] Determine whether the first deep learning model classifies the input speech as real speech (step S613). When the input speech is not real speech, reject the false speech (step S614). When the input speech is real speech, use a second deep learning model to process the input speech to generate a speech feature (step S615).

[0070] Next, compare the voice feature with at least one preset feature to generate a comparison result (step S616). In one possible embodiment, the comparison result is the similarity between the voice feature and the preset feature. Then, determine whether the comparison result is greater than a threshold (step S617). When the comparison result is not greater than the threshold, it means that the input voice is not the voice of a legitimate person. Therefore, no specific action is performed (step S618). When the comparison result is greater than the threshold, it means that the input voice is the voice of a legitimate person. Therefore, a specific action is performed (step S619).

[0071] The present invention does not limit the type of specific action. In one possible embodiment, the specific action opens the door lock. In another possible embodiment, when the comparison result is not greater than the threshold, it indicates that an illegal person attempts to enter the house, so step S618 emits a warning sound to remind the person in the house. In other embodiments, the strictness of voice recognition can be improved by adjusting the threshold.

[0072] The speech recognition method of the present invention, or a specific form or part thereof, can exist in the form of program code. The program code can be stored in a physical medium, such as a floppy disk, a CD, a hard disk, or any other machine-readable (such as computer-readable) storage medium, or a computer program product not limited to an external form, wherein when the program code is loaded and executed by a machine, such as a computer, the machine becomes used to participate in the attack defense module and speaker verification module of the present invention. The program code can also be transmitted through some transmission media, such as wires or cables, optical fibers, or any transmission mode, wherein when the program code is received, loaded and executed by a machine, such as a computer, the machine becomes used to participate in the attack defense module and speaker verification module of the present invention. When implemented in a general-purpose processing unit, the program code combines with the processing unit to provide a unique device that operates similarly to an application-specific logic circuit.

[0073] Unless otherwise defined, all words (including technical and scientific words) herein are generally understood by those skilled in the art to which the present invention belongs. In addition, unless expressly indicated, the definitions of words in general dictionaries should be interpreted as consistent with the meanings in articles in the relevant technical field, and should not be interpreted as ideal or overly formal. Although the terms "first", "second", etc. can be used to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element.

[0074] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art may make some changes and modifications without departing from the spirit and scope of the present invention. For example, the system, device or method described in the embodiments of the present invention may be implemented in a physical embodiment of hardware, software or a combination of hardware and software. Therefore, the scope of protection of the present invention shall be subject to the scope defined by the attached claims.

Claims

1. A speech recognition device, characterized in that: include: An attack defense module is used to filter out a false voice, and includes: a first processing circuit for processing a first input voice to generate a first audio signal; and an inference circuit, receiving the first audio signal and processing the first audio signal using a first deep learning model to determine whether the first input speech is a real speech or the false speech, and when the first input speech is the real speech, the inference circuit outputs a second input speech; and One-speaker verification module, including: a feature extraction circuit, processing the second input speech using a second deep learning model to generate a speech feature; a comparison circuit, for comparing the voice feature with at least one preset feature to generate a comparison result; and A decision circuit determines whether the comparison result is greater than a threshold. When the comparison result is greater than the threshold, the decision circuit performs a specific action. When the comparison result is not greater than the threshold, the decision circuit does not perform the specific action.

2. The speech recognition device according to claim 1, wherein: When the first input voice is the real voice, the inference circuit uses the first audio signal as the second input voice.

3. The speech recognition device according to claim 1, wherein: When the first input voice is the true voice, the inference circuit uses the first input voice as the second input voice.

4. The speech recognition device according to claim 3, wherein: The speaker verification module also includes: a second processing circuit for processing the second input voice to generate a second audio signal; The feature extraction circuit receives the second audio signal and processes the second audio signal using the second deep learning model to generate the speech feature.

5. The speech recognition device according to claim 4, wherein: During a registration period, the second processing circuit processes a third input speech to generate a third audio signal, the feature extraction circuit receives the third audio signal, and processes the third audio signal using the second deep learning model to generate the preset feature.

6. The speech recognition device according to claim 5, wherein: The speaker verification module also includes: A storage circuit is used to store the preset characteristic.

7. The speech recognition device according to claim 1, wherein: The attack defense module also includes: A first input / output interface, used for receiving a first update data; During a first update period, the inference circuit updates at least one parameter of the first deep learning model according to the first update data.

8. The speech recognition device according to claim 1, wherein: The speaker verification module also includes: A second input / output interface, used for receiving a second update data; During a second update period, the feature extraction circuit updates at least one parameter of the second deep learning model according to the second update data.

9. The speech recognition device according to claim 1, wherein: Also includes: A receiving circuit is used to receive the first input voice and provide the first input voice to the attack defense module.

10. A speech recognition method, characterized in that: include: Receive and process an input voice to generate an audio signal; Processing the audio signal using a first deep learning model to determine whether the input speech is a real speech or a fake speech; When the input speech is the real speech, using a second deep learning model to process the input speech to generate a speech feature; comparing the speech feature with at least one predetermined feature to generate a comparison result; Determine whether the comparison result is greater than a threshold; When the comparison result is greater than the threshold, performing a specific action; and When the comparison result is not greater than the threshold, the specific action is not performed.

Images