Control and procedure for activating a vehicle function by authenticating a voice sample

Abstract

The controller 100 detects the voice command using at least one sensor 102 installed in the vehicle. The controller 100 filters the speech sample from the voice command using the speech activity detection module 104. The controller 100 checks the speech sample for a recorded or imitated voice using the deception detection module 106, extracts the speech signature from the speech sample using the speech signature extractor 108, and authenticates the speech signature using an authenticator 110 by referencing the pre-recorded speech signature from the user login process. From the speech signature, the controller 100 extracts the entity and intent using the entity and intent extractor 112 and validates the entity and intent using the validation module 114 by referencing pre-configured access data for multiple users simultaneously.After validation, control unit 100 executes the intention to activate the vehicle functions.

Description

Field of invention The present invention relates to a control system and a method for activating a vehicle function by authenticating a voice sample. Background of the invention In vehicles, authenticating vehicle functions with level-based access control allows all registered users to operate them, which can lead to unintended actions. After registration, the user gains access to all vehicle functions, including sensitive information belonging to other users or the vehicle owner. This poses a risk in multi-user environments, such as a vehicle, where anyone can access sensitive functions intended for a specific user, like the vehicle owner, raising security concerns. These existing methods require a stable internet connection for user authentication, which can lead to authentication failures in areas with weak internet connectivity. Storing sensitive user data in the cloud raises privacy and security concerns. According to the prior art (US2019043295), a microcontroller comprises a processor and instructions stored on a machine-readable medium. The instructions, which are loaded and executed by the processor, cause the processor to detect a transmitter's access to a vehicle, recognize a voice command to open a vehicle part, and, based on the combination of voice command recognition and transmitter access detection, output a signal to activate a vehicle function identified in the voice command. Brief description of the enclosed drawings: One embodiment of the disclosure is described with reference to the accompanying drawings below. Fig. 1 shows a block diagram of a control system for activating a vehicle function by authenticating a voice sample according to one embodiment of the present invention, and Fig. 2 shows various steps that are performed during the process of activating a vehicle function by authenticating a voice sample according to one embodiment of the present invention. Detailed description of embodiments Fig. 1 shows a block diagram of a controller 100 for activating a vehicle function by authenticating a voice sample according to an embodiment of the present invention. The controller 100 is configured to detect a voice command by means of at least one sensor 102 installed in the vehicle. The controller 100 is further configured to filter a voice sample from the voice command by means of a voice activity detection module 104. The controller 100 is configured to check the voice sample for a recorded or imitated voice by means of a deception detection module 106. The controller 100 is configured to extract a voice signature from the voice sample by means of a voice signature extractor 108 and to authenticate the voice signature by means of an authenticator 110 by reference to a previously recorded voice signature from a user login process.The controller 100 is further configured to extract an entity and an intention from the speech signature using an entity and intention extractor 112, characterized in that the controller 100 is configured to validate the entity and the intention using a validation module 114 by reference to preconfigured access data, wherein the entity and the intention are validated simultaneously for multiple users with respect to their speech signatures. After validating the entity and the intention, the controller 100 is configured to execute the intention to activate the vehicle function. According to one embodiment of the present invention, the controller 100 measures the duration of the speech sample filtered out by the speech activity recognition module 104 to activate the vehicle function by authenticating the speech sample. The controller 100 processes the filtered speech sample to generate a speech instance using a buffer memory module 116 if the duration of the speech sample is shorter than a threshold duration. The threshold duration represents the minimum duration of the speech sample required for speech evaluation. The buffer memory module 116 is only used if the duration of the speech sample is shorter than the threshold duration. However, if the duration of the speech sample equals or exceeds the threshold duration, the speech sample is processed directly, and the buffer memory module 116 is not required. According to one embodiment of the present invention, the buffer storage module 116 is an extension module that uses at least one method from a group consisting of random sectioning and random rearrangement, sectioning into overlapping window sections, and speech activity detection and rearrangement to generate the speech instance from the speech sample. The random sectioning method comprises slicing the speech sample into several small sections and then randomly rearranging these sections, thereby generating the speech instance. This preserves phase information of the speech sample, thus maintaining the naturalness of the speech sample in the speech instance. The energy-weighted frame scaling method comprises processing the speech sample by segmenting it into individual frames.Each frame is analyzed to calculate its energy level, and a weighting procedure is applied based on this calculated energy level. Frames with higher energy levels are prioritized and inserted first into the final sequence, resulting in the speech instance. The speech activity detection and rearrangement process identifies and isolates individual words or phonetic units within the speech sample. Once identified and isolated, these units are randomly rearranged to generate the speech instance. According to the embodiment of the present invention, the user login process is performed by the controller 100. The controller 100 detects the voice command using at least one sensor 102 installed in the vehicle. The controller 100 identifies the voice sample in the voice command using the voice activity detection module 104. After identifying the voice sample, the controller 100 checks the voice sample for a recorded or imitated voice using the deception detection module 106. The controller 100 extracts the voice signature from the voice sample using the voice signature extractor 108 and verifies the voice signature against a previously existing voice signature. This verification ensures that the user does not attempt to log in multiple times and prevents duplicate logins. After verification, the controller 100 records the voice signature and assigns the desired vehicle functions to the voice signature.Vehicle functions are selected from a group that includes adjusting the seat temperature, changing a song, changing a radio station, and the like. Once various vehicle functions are assigned to the voice signature, the user is logged in. The Control 100 avoids duplicate logins by using a signature recording module. According to the embodiment of the present invention, the entity and intent extractor 112 is a machine learning-based intent / entity extraction algorithm based on natural language processing, and is a supervised algorithm. The algorithm is trained using voice commands selected from a group that includes playing a track, changing the track, changing a radio station, activating an air conditioner, and the like. The controller 100 validates the extracted entity and intent by referencing pre-configured access data for multiple users simultaneously. After validation, the controller 100 executes the intent to activate the vehicle functions. According to one embodiment of the present invention, the controller 100 is equipped with the necessary circuit arrangements for signal acquisition, recording, and processing. The controller 100 has an input interface, output interfaces with pins or ports, a memory element such as random access memory (RAM) and / or read-only memory (ROM), an analog-to-digital converter (ADC) and a digital-to-analog converter (DAC), a clock, a timer, a counter, and at least one processor (capable of implementing machine learning), which are interconnected with each other and with other components via communication bus systems. The memory element (not shown) contains logic, instructions, programs, applications, or modules / models and / or thresholds.The controller 100 stores predefined areas, reference values, predefined / predetermined criteria / conditions, lists, and knowledge sources that the at least one processor accesses according to the defined procedures. The internal components of the controller 100 are not described in detail, as they are state-of-the-art and are not to be understood as limiting. The controller 100 may also include communication units such as transceivers for communication via wireless or wired transmission paths, such as GSM (Global System for Mobile Communications), 3G, 4G, 5G, Wi-Fi, Bluetooth, Ethernet, serial networks, and the like. The controller 100 can be implemented as a system-in-package (SiP), a system-on-chip (SoC), or in any other known embodiment. Examples of the controller 100 include, but are not limited to, microcontrollers, microprocessors, microcomputers, electronic control units (ECUs), etc. According to the embodiment of the present invention, the control unit 100 is operated. The control unit 100 activates the vehicle function by authenticating the voice sample. The control unit 100 detects the voice command by means of at least one sensor 102 installed in the vehicle. The voice command is detected by sensors mounted in the vehicle, for example, by microphones located in the vehicle, such as on the dashboard, the mirror, etc. Once the voice command has been detected, the control unit 100 identifies the voice sample in the voice command by means of the speech activity recognition module 104. The speech activity recognition module 104 uses methods such as a fast Fourier transform (FFT), a zero-crossing rate (ZCR), etc., to determine the energy and frequency of the voice command, based on which the control unit 100 identifies the voice sample, background noise, and non-speech components.The speech activity detection module 104 filters out background noise and non-speech components and identifies the speech sample, which is then used by the controller 100 for further processing. The controller 100 checks the speech sample for recorded or imitated voices using a deception detection module 106. The deception detection module 106 is a supervised machine learning model that protects against unauthorized access by identifying and filtering out recorded or imitated speech samples. The controller 100 extracts the speech signature from the speech sample using the speech signature extractor 108 and authenticates the speech signature using the authenticator 110 by referencing the previously recorded speech signature from the user login process. During the user login process, the speech signature is extracted from the speech sample and recorded.The controller 100 accesses this pre-recorded voice signature and authenticates the user. From the voice sample, the controller 100 further extracts the entity and intent using the entity and intent extractor 112 and validates the entity and intent using the validation module 114 by referencing pre-configured access data. During user login, the voice signature of the logged-in user with access to various vehicle functions is pre-configured and stored in a configurator, which the controller 100 references during the validation of the entity and intent. The controller 100 simultaneously validates the entity and intent for multiple users with respect to their respective voice signatures and executes the intent to activate the vehicle functions. According to the present invention, the various modules used by the controller 100 to activate the vehicle function by authenticating the voice sample are the voice activity detection module 104, the deception detection module 106, the voice signature extractor 108, the framework on which the entity and intent extractor 112 is based, the authenticator 110, and the signature recording module. These modules can be developed independently, created internally, or adopted from open-source sources available in the public domain. The data set required for training these modules can also be developed independently, created internally, or adopted from open-source sources available in the public domain. According to the present invention, the controller 100 and the method for activating a vehicle function by authenticating a voice sample are disclosed. The controller 100 activates the vehicle function by authenticating a voice sample. The voice-based activation allows all logged-in users to use only those functions for which they are authorized. Otherwise, a user would have unrestricted access to all functions after logging in, which poses a security risk. The voice evaluation, in combination with the classification of entity and intent, enables different access levels based on a configurator. The configurator stores voice signatures of several users, each assigned different vehicle functions, which the controller accesses during authentication. This enables different access levels for different user types.For example, the system restricts 100 guests from performing safety-critical functions that are reserved for regular users or the vehicle owner. Simultaneous interaction by multiple users allows for independent control of various vehicle functions, such as the driver changing the driving mode, a passenger adjusting the temperature, and so on. This independent control enables a dynamic and user-centric experience. According to the present invention, the steps for activating the vehicle function by authenticating the voice sample are performed locally by the controller 100, without dependence on an available internet connection. For example, these steps include capturing the voice command, identifying the voice sample within the voice command, checking the voice sample against a recorded or imitated voice using the deception detection module 106, extracting the voice signature from the voice sample, extracting the entity and intent from the authenticated voice signature, validating the entity and intent using the validation module 114, and executing the intent to activate the vehicle functions. None of these steps are performed on a server or in a cloud environment.In a cloud-based system, user authentication depends on a stable internet connection, which can lead to authentication failures in areas with poor connectivity. Storing sensitive user data, such as voice samples, in the cloud raises privacy concerns. Unauthorized access to the cloud environment can compromise voice samples and lead to significant security and privacy breaches. According to one embodiment of the present invention, the controller 100 registers the user with a voice signature, wherein the voice signature is associated with one or more vehicle functions. After registration is complete, the registered user is authenticated during real-time operation, and access to the authorized vehicle functions is granted. For example, a vehicle owner registers their voice signature for functions such as changing the music track, setting the air conditioning temperature to a specific value, opening the trunk, etc. During authentication, the voice signature is extracted when the owner utters the voice command. Once the voice signature is authenticated, the intent and entity are extracted. In this context, the intent and entity provide information about the owner and the vehicle function the owner wishes to access.Once it has been validated that the owner has access to the desired function, the controller 100 performs the corresponding action. According to the embodiment of the present invention, the controller 100 performs the authentication of multiple users simultaneously. For example, in a vehicle with two users, i.e., User 1 and User 2, both users simultaneously enter voice commands to access a vehicle function. Based on the authentication, the controller 100, provided both User 1 and User 2 are registered, executes the respective commands and activates the associated functions. This concept of voice recognition can also be transferred to other applications, for example, a coffee machine with a personalized user profile. Multiple users register with individual settings. Upon receiving the voice command, the coffee machine identifies the registered user and activates the corresponding presets. This enables personalized adaptation to individual needs in a multi-user environment.The example above describes one embodiment of the speech recognition concept among many possible embodiments that may arise from the present disclosure. It is therefore not to be interpreted restrictively. Fig. 2 shows various steps performed in the course of the method for activating vehicle functions by authenticating the voice sample according to the present invention. The method comprises a plurality of steps, wherein step 202 includes the detection of the voice command by the controller 100 using at least one sensor 102 installed in the vehicle. Step 204 includes the filtering out of the voice sample in the voice command by the controller 100 using the voice activity recognition module 104. Step 206 includes the checking of the voice sample with regard to a recorded or imitated voice by the controller 100 using the deception detection module 106.Step 208 comprises the extraction of the speech signature from the speech sample by the controller 100 using the speech signature extractor 108 and the authentication of the speech signature using the authenticator 110 by referencing the pre-recorded speech signature from the user login process. Step 210 comprises the extraction of the entity and intent from the speech signature by the controller 100 using the entity and intent extractor 112. The procedure is characterized by a step 212 that includes the validation of the entity and intent by the controller 100 using the validation module 114 by referencing pre-configured access data. The entity and intent are validated simultaneously for multiple users with respect to their respective speech signatures. Step 214 comprises the execution of the intent by the controller 100 to activate the vehicle function. According to the procedure, activating the vehicle function by authenticating the speech sample further includes a step 216, which involves the controller 100 measuring the duration of the speech sample filtered out by the speech activity recognition module 104. A step 218 involves the controller 100 processing the filtered speech sample to generate a speech instance using the buffer memory module 116, provided the duration of the speech sample is shorter than the threshold duration. The buffer memory module 116 processes the speech sample only if its duration is shorter than the threshold duration. The duration of the speech instance is longer than the duration of the speech sample.The buffer storage module 116 is an extension model that uses at least one method from a group that includes random sectioning and random rearrangement, sectioning into overlapping window sections, and speech activity detection and rearrangement to generate the speech instance from the speech sample. According to the procedure, this further includes user login by the controller 100. The procedure includes the controller 100 capturing the voice command by means of at least one sensor 102 installed in the vehicle. The procedure includes the controller 100 filtering out the speech sample in the voice command by means of the speech activity detection module 104. The procedure includes the controller 100 checking the speech sample for a recorded or imitated voice by means of the deception detection module 106. The procedure includes the controller 100 extracting the speech signature from the speech sample and verifying the speech signature against a previously existing speech signature.The procedure involves recording voice signatures for multiple users via the Control Unit 100 and assigning these voice signatures to different vehicle functions. These functions are selected from a group and include adjusting the seat temperature, changing a song, changing a radio station, and the like. The procedure also includes logging in multiple users via the Control Unit 100. Access to multiple vehicle functions is assigned to a single voice signature. For example, the vehicle owner has access to all vehicle functions, a relative of the owner has access to five functions, and the vehicle driver has access to four functions. It is possible for the owner and the driver to have access to some of the same functions. Access to these functions is customizable and can be tailored to specific requirements. It is understood that the embodiments described above are merely examples and do not limit the scope of the invention. Numerous such embodiments, as well as further modifications and changes to the embodiment described above, are provided for. The scope of the invention is limited exclusively by the scope of the claims. QUOTES INCLUDED IN THE DESCRIPTION This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature US 2019043295

[0003]

Claims

A controller (100) for activating a vehicle function by authenticating a voice sample, wherein the controller (100) is configured to: a. detect a voice command by means of at least one sensor (102) installed in the vehicle; b. extract a voice sample from the voice command by means of a voice activity detection module (104); c. check the voice sample with regard to a recorded or imitated voice by means of a deception detection module (106); d. extract a voice signature from the voice sample by means of a voice signature extractor (108) and authenticate the voice signature by means of an authenticator (110) with reference to a previously recorded voice signature from a user login process; e. extract an entity and an intent from the voice signature by means of an entity and intent extractor (112), characterized by f.Validating the entity and intent using a validation module (114) with reference to preconfigured access data, wherein the entity and intent are validated simultaneously for multiple users with respect to their respective language signatures, and executing the intent to activate the vehicle function. Control (100) for activating a vehicle function by authenticating a speech sample according to claim 1, further comprising: a. measuring the duration of the speech sample filtered out by the speech activity recognition module (104), and b. processing the filtered speech sample to generate a speech instance using a buffer memory module (116), if the duration of the speech sample is shorter than the threshold duration. Controller (100) according to claim 2, wherein the buffer storage module (116) uses at least one process from a group comprising: a. random splitting into sections and random rearrangement, b. splitting into overlapping window sections, and c. speech activity detection and rearrangement. Control unit (100) according to claim 1, wherein the user login comprises the following steps: a. capturing the voice command by means of the at least one sensor (102) installed in the vehicle; b. filtering the speech sample from the voice command by the speech activity detection module (104); c. checking the speech sample with regard to a recorded or imitated voice by the deception detection module (106); d. extracting the speech signature from the speech sample by means of the speech signature extractor (108) and checking the speech signature against a previously existing speech signature; and e. recording the speech signature for the respective multiple users and assigning the speech signature to different vehicle functions, wherein the vehicle functions are selected from a group that includes adjusting the seat temperature, changing a song, changing a radio station, and the like. Controller (100) according to claim 1, wherein the entity and intent extractor (112) is a natural language processing-based model, wherein the model is trained by means of a voice command from a group comprising playing a track, changing the track, changing a radio station, activating an air conditioner and the like. A method for activating a vehicle function by authenticating a voice sample, wherein the method is performed by a controller (100), the method comprising the steps: a. capturing a voice command by means of at least one sensor (102) installed in the vehicle, b. filtering out a voice sample from the voice command by means of a voice activity detection module (104), c. checking the voice sample with regard to a recorded or imitated voice by means of a deception detection module (106), d. extracting a voice signature from the voice sample by means of a voice signature extractor (108) and authenticating the voice signature by means of an authenticator (110) by reference to a pre-recorded voice signature from a user login process, e. extracting an entity and an intent from the voice signature by means of an entity and intent extractor (112), characterized by f.Validating the entity and intent using a validation module (114) by reference to preconfigured access data, wherein the entity and intent are validated simultaneously for multiple users with respect to their respective language signatures, and executing the intent to activate the vehicle function. Method for activating the vehicle function by authenticating the speech sample, wherein the method is carried out by the controller (100), the method further comprising the steps: a. measuring a duration of the speech sample filtered out by the speech activity recognition module (104), and b. processing the filtered speech sample to generate a speech instance using a buffer storage module (116), if the duration of the speech sample is shorter than the threshold duration. Method according to claim 7, wherein the buffer storage module (116) uses at least one process from a group comprising: a. random splitting into sections and random rearrangement, b. splitting into overlapping window sections, and c. speech activity detection and rearrangement. The method of claim 6, wherein the registration process, performed by the controller (100), comprises the following steps: a. capturing the voice command by means of the at least one sensor (102) installed in the vehicle, b. filtering the speech sample from the voice command by the speech activity detection module (104), c. checking the speech sample with regard to a recorded or imitated voice by the deception detection module (106), d. extracting the speech signature from the speech sample by means of the speech signature extractor (108) and checking the speech signature against a previously existing speech signature, and e. recording the speech signature for multiple users and assigning the speech signature to different vehicle functions, wherein the vehicle functions are selected from a group that includes adjusting the seat temperature, changing a song, changing a radio station, and the like. The method of claim 6, wherein the entity and intent extractor (112) is a natural language processing-based model, wherein the model is trained by means of a speech command from a group comprising playing a track, changing the track, changing a radio station, activating an air conditioner and the like.

Images