CN112735457B - Voice denoising method and system - Google Patents
Voice denoising method and system Download PDFInfo
- Publication number
- CN112735457B CN112735457B CN202011449222.3A CN202011449222A CN112735457B CN 112735457 B CN112735457 B CN 112735457B CN 202011449222 A CN202011449222 A CN 202011449222A CN 112735457 B CN112735457 B CN 112735457B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- target
- noise reduction
- reduction algorithm
- model library
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 230000009467 reduction Effects 0.000 claims abstract description 128
- 230000000694 effects Effects 0.000 claims abstract description 15
- 230000008030 elimination Effects 0.000 claims abstract description 11
- 238000003379 elimination reaction Methods 0.000 claims abstract description 11
- 230000008569 process Effects 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000004891 communication Methods 0.000 claims description 21
- 238000001514 detection method Methods 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000009191 jumping Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Navigation (AREA)
- Traffic Control Systems (AREA)
Abstract
The embodiment of the application discloses a voice denoising method and system, and relates to the technical field of Internet of vehicles and voice processing. The method comprises the following steps: collecting noise in the driving process of the vehicle, and extracting characteristic information of the noise; if the target noise reduction algorithm matched with the characteristic information does not exist in the offline model library of the vehicle, determining a target vehicle matched with the driving environment and the environment in the vehicle, and requesting the target noise reduction algorithm from the offline model library of the target vehicle; and if the target noise reduction algorithm is received, denoising the voice of the user of the vehicle by adopting the target noise reduction algorithm, and adding the target noise reduction algorithm into the offline model library. In the embodiment, when the target noise reduction algorithm does not exist in the vehicle, the target noise reduction algorithm is requested from the target vehicle matched with the driving environment and the environment in the vehicle, so that the noise reduction algorithm is selected and obtained according to the driving environment and the environment in the vehicle, and the noise elimination effect is improved.
Description
Technical Field
The embodiment of the application relates to the technologies of Internet of vehicles and voice processing, in particular to a voice denoising method and system.
Background
With the development of the technology of the automatic driving automobile, drivers are gradually freed from the driving task. The man-machine interaction mode is greatly changed, and the voice interaction is widely applied due to the unique convenience.
Because the driving scene of the automatic driving vehicle is complex, the environmental noise has great difference according to the different driving scenes, so that a series of problems of low recognition rate, high hardware cost, poor elimination effect of the environmental noise of the automatic driving vehicle and the like exist in the current voice interaction, and the vehicle can possibly fail to correctly understand the voice command of the passenger.
Disclosure of Invention
The embodiment of the application provides a voice denoising method and system, so that a denoising algorithm is selected and obtained according to scenes inside and outside a vehicle, and the noise elimination effect is improved.
In a first aspect, an embodiment of the present application provides a speech denoising method, which is applicable to a vehicle-mounted terminal, and includes:
collecting noise in the driving process of the vehicle, and extracting characteristic information of the noise;
if the target noise reduction algorithm matched with the characteristic information does not exist in the offline model library of the vehicle, determining a target vehicle matched with the driving environment and the environment in the vehicle, and requesting the target noise reduction algorithm from the offline model library of the target vehicle;
if a target noise reduction algorithm in an offline model library of the target vehicle is received, denoising the voice of the user of the vehicle by adopting the target noise reduction algorithm, and adding the target noise reduction algorithm into the offline model library;
the driving environment comprises at least one of a POI (point of interest), road condition information, road section information and vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
In a second aspect, an embodiment of the present application further provides a speech denoising system, including: the vehicle, the target vehicle and the cloud;
the vehicle is used for collecting noise in the driving process of the vehicle and extracting the characteristic information of the noise; if the target noise reduction algorithm matched with the characteristic information does not exist in the off-line model library of the vehicle, determining the driving environment of the vehicle according to the acquisition information of the sensor and the road condition information; sending the driving environment and the environment in the vehicle to a cloud end; receiving the target noise reduction algorithm from the cloud; denoising the voice of the user of the vehicle by adopting the target denoising algorithm, and adding the target denoising algorithm into the offline model library;
the cloud is used for determining a target vehicle matched with the running environment and the in-vehicle environment of the vehicle, requesting the target noise reduction algorithm from the offline model library of the target vehicle and returning the target noise reduction algorithm to the offline model library of the vehicle;
the driving environment comprises at least one of a POI (point of interest), road condition information, road section information and vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
In a third aspect, an embodiment of the present application further provides a speech denoising system, including: the vehicle-mounted system comprises a host vehicle and at least one candidate vehicle within a set distance range with the host vehicle;
the vehicle is used for collecting noise in the driving process of the vehicle and extracting the characteristic information of the noise; if the target noise reduction algorithm matched with the characteristic information does not exist in the off-line model library of the vehicle, establishing communication connection with the candidate vehicle; acquiring the in-vehicle environment of the candidate vehicles based on the communication connection, and determining a target vehicle matched with the in-vehicle environment and the vehicle from the candidate vehicles; requesting the target noise reduction algorithm from an offline model library of the target vehicle based on a communication connection with the target vehicle; if the target noise reduction algorithm sent by the target vehicle is received, denoising the voice of the user of the vehicle by adopting the target noise reduction algorithm, and adding the target noise reduction algorithm into the offline model library;
the candidate vehicle is used for establishing communication connection with the vehicle and sending the environment in the vehicle to the vehicle based on the communication connection, and the target noise reduction algorithm;
the driving environment comprises at least one of a POI (point of interest), road condition information, road section information and vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
In the embodiment of the application, the target noise reduction algorithm and the characteristic information are stored in the offline model library, so that vehicle computing resources are saved, the running speed is increased, the vehicle computing resources are saved, and the running speed is increased; moreover, through the matching of the algorithm and the characteristic information, the existing algorithm can be directly called for noise elimination, the noise reduction sound acquisition equipment of the vehicle is saved, and the cost is reduced; when the target noise reduction algorithm matched with the characteristic information does not exist in the off-line model base of the vehicle, the target noise reduction algorithm is requested from the off-line model base of the target vehicle matched with the driving environment and the in-vehicle environment of the vehicle, so that the noise reduction algorithm is selected and obtained according to the characteristic that the noise is similar when the driving environment is matched with the in-vehicle environment, and the noise elimination effect is improved.
Drawings
Fig. 1 is a flowchart of a first speech denoising method according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a speech denoising system according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of another speech denoising system according to an embodiment of the present application.
Detailed Description
The present application will be described in further detail with reference to the following drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the application and are not limiting of the application. It should be further noted that, for the convenience of description, only some of the structures related to the present application are shown in the drawings, not all of the structures.
The embodiment of the application provides a first voice denoising method, a flow chart of which is shown in fig. 1, and which is applicable to denoising voice of a user in a vehicle during driving of the vehicle. The method may be performed by a speech denoising apparatus, which may be constituted by software and/or hardware, and generally integrated in a vehicle-mounted terminal.
With reference to fig. 1, the method provided in this embodiment specifically includes:
s110, collecting noise in the driving process of the vehicle, and extracting characteristic information of the noise.
Specifically, the noise in the driving process of the vehicle is collected through scene noise collecting equipment. The noise may be caused by the running environment or the environment inside the vehicle.
Then, the acquired noise is subjected to data processing, and characteristic information of the noise, such as spectral characteristics, is extracted.
Optionally, the scene noise collection device transmits the noise to the feature extraction module through the vehicle-mounted network or the bluetooth module, so as to extract feature information of the noise through the feature extraction module.
S120, judging whether a target noise reduction algorithm matched with the characteristic information exists in the offline model library of the vehicle, and if so, jumping to S130; if not, it jumps to S140.
The vehicle and the target vehicle are both provided with an off-line model library, and the off-line model library stores the corresponding relation between the characteristic information and the noise reduction algorithm. The construction process in the off-line model library is described in detail below.
The first step is as follows: and acquiring noises in different driving environments and in-vehicle environments through scene noise acquisition equipment.
The scene noise acquisition equipment is specifically a vehicle-mounted noise acquisition module, and is used for storing acquired noise into a scene noise acquisition and storage module and uploading the acquired noise to an original scene storage module of a scene noise data management system.
The second step is that: and carrying out data processing on each collected noise, and extracting characteristic information. Optionally, the characteristic information of the noise is obtained by connecting wavelet transform.
The third step: classifying the extracted characteristic information of the noise and constructing an offline scene noise library; and storing the noise reduction algorithm corresponding to each type of feature information in an offline noise reduction algorithm library. That is, the offline scene noise library and the offline noise reduction algorithm library together constitute an offline model library.
Optionally, the determining whether a target noise reduction algorithm matching the feature information exists in the offline model library of the vehicle includes: and comparing the characteristic information in an off-line scene noise library. And if the matched characteristic information exists in the off-line scene noise library, denoising the voice of the user of the vehicle through a corresponding off-line denoising algorithm. The denoised speech may then be responded to execute the user instructions.
S130, denoising the voice of the user by adopting the target denoising algorithm. And finishing the operation.
S140, determining a target vehicle matched with the running environment and the environment in the vehicle, and requesting the target noise reduction algorithm from the off-line model library of the target vehicle. Execution continues with S150.
Specifically, the driving environment includes at least one of a Point of Information (POI), road condition Information, road segment Information, and a vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
The noise of the vehicle is different when the vehicle is positioned at different POIs, such as a parking lot, a hospital, a school and a park, the noise is different when the vehicle is in traffic jam, bumpiness, smoothness and other road conditions, the noise is different when the vehicle is on different road sections such as a high speed road, a rural road, a national road and the like, and the noise is different when the vehicle speed is different. On the other hand, when the model of the engine of the vehicle, the cold start state, the opening and closing of the vehicle window, whether audio and video are played or not, the volume and the like are different, the noise is also obviously different. However, the vehicle noise of the same running environment and in-vehicle environment is similar, and therefore, an appropriate target vehicle can be selected depending on the running environment and in-vehicle environment. And a target noise reduction model is being used for noise reduction on the target vehicle, or even if the noise reduction is not performed, a target noise reduction algorithm is probably included in an offline model library of the target vehicle.
The specific implementation of requesting the target noise reduction algorithm from the off-line model library of the target vehicle will be described in the following examples.
S150, judging whether a target noise reduction algorithm in the target vehicle offline model library is received. If yes, jumping to S160; if not, it jumps to S170.
S160, denoising the voice of the user of the vehicle by adopting the target denoising algorithm, and adding the target denoising algorithm into the offline model library. And finishing the operation.
And if the target noise reduction algorithm exists in the offline model library of the target vehicle, the host vehicle receives the target noise reduction algorithm sent by the target vehicle. And updating the off-line model library while denoising.
S170, requesting the target noise reduction algorithm from the cloud. The target noise reduction algorithm is stored in a cloud end in advance, or the target noise reduction algorithm is obtained by carrying out noise elimination on the characteristic information of the noise through the cloud end.
And if the target noise reduction algorithm does not exist in the offline model base of the target vehicle, the target noise reduction algorithm in the offline model base of the target vehicle is not received. Then, in order to ensure smooth noise reduction processing, the vehicle uploads the collected characteristic information and the positioning information of the noise to the cloud through the cloud uploading module. Furthermore, the vehicle also requests the positioning information of the target vehicle and uploads the positioning information of the target vehicle to the cloud.
The cloud compares the collected characteristic information of the noise in a cloud scene noise library, wherein the cloud scene noise library stores characteristic information of various noises. And if the matched characteristic information exists, returning the corresponding target noise reduction algorithm to the off-line model base of the vehicle and the target vehicle according to the positioning information of the vehicle and the target vehicle. If the matched characteristic information does not exist, the cloud end carries out noise elimination on the characteristic information of the noise to obtain a target noise reduction algorithm; and returning the target noise reduction algorithm to the off-line model libraries of the vehicle and the target vehicle according to the positioning information of the vehicle and the target vehicle.
S180, denoising the voice of the vehicle user by adopting the target denoising algorithm, and adding the target denoising algorithm into an offline model library of the vehicle and the target vehicle through the cloud.
In the embodiment of the application, the target noise reduction algorithm and the characteristic information are stored in the offline model library, so that vehicle computing resources are saved, the running speed is increased, the vehicle computing resources are saved, and the running speed is increased; moreover, through the matching of the algorithm and the characteristic information, the existing algorithm can be directly called for noise elimination, the noise reduction sound acquisition equipment of the vehicle is saved, and the cost is reduced; when the target noise reduction algorithm matched with the characteristic information does not exist in the off-line model base of the vehicle, the target noise reduction algorithm is requested from the off-line model base of the target vehicle matched with the driving environment and the in-vehicle environment of the vehicle, so that the noise reduction algorithm is selected and obtained according to the characteristic that the noise is similar when the driving environment is matched with the in-vehicle environment, and the noise elimination effect is improved.
In the above-described embodiment and the following embodiments, the target vehicle matching the driving environment and the in-vehicle environment of the host vehicle is determined, and the target noise reduction algorithm is requested from the offline model library of the target vehicle, including the following two alternative implementations.
First optional implementation (determined by cloud): determining the driving environment of the vehicle according to the acquisition information and the road condition information of the sensor; and sending the running environment and the in-vehicle environment of the vehicle to a cloud end, so that the cloud end can determine a target vehicle matched with the running environment and the in-vehicle environment of the vehicle, request the target noise reduction algorithm from an offline model library of the target vehicle, and return the target noise reduction algorithm to the offline model library of the vehicle.
Wherein, the sensor includes but is not limited to a camera, a vehicle radar and an inertial navigation system. The road condition information can be obtained according to the vehicle-mounted high-precision map, such as congestion, bump, smoothness and the like. And comprehensively acquiring the information and the road condition information to obtain the driving environment of the vehicle. The environment in the vehicle CAN be obtained by reading a corresponding signal from a Controller Area Network (CAN) bus.
The cloud end obtains the running environment and the in-vehicle environment of all vehicles in real time, determines a target vehicle matched with the running environment and the in-vehicle environment of the vehicle, requests the target noise reduction algorithm from the off-line model library of the target vehicle, and returns the target noise reduction algorithm to the off-line model library of the vehicle.
Second alternative implementation (point-to-point determination): searching at least one candidate vehicle within a set distance range with the vehicle, and establishing communication connection with each candidate vehicle; acquiring the in-vehicle environment of each candidate vehicle based on the communication connection, and determining a target vehicle matched with the in-vehicle environment and the vehicle from each candidate vehicle; requesting the target noise reduction algorithm from an offline model library of the target vehicle based on a communication connection with the target vehicle.
The set distance range may be determined according to a distance range of a communication connection, which is not limited to WIFI and bluetooth, for example, 10 meters or 20 meters. For convenience of description and distinction, the vehicles within the own vehicle set distance range are referred to as candidate vehicles. Due to the close distances, the driving environment of the candidate vehicle is similar to that of the host vehicle. Based on this, the driving environment does not need to be matched.
After establishing a communication connection with each candidate vehicle, request information of the in-vehicle environment is transmitted to each candidate vehicle based on the communication connection. Each candidate vehicle obtains the in-vehicle environment from the CAN signal in response to the request information, and returns to the own vehicle based on the communication connection. The vehicle determines a target vehicle matched with the vehicle in the vehicle environment from the candidate vehicles; requesting the target noise reduction algorithm from an offline model library of the target vehicle based on a communication connection with the target vehicle.
In the foregoing embodiment and the following embodiments, after the denoising processing is performed on the voice of the host vehicle user by using the target denoising algorithm, the method further includes: determining the noise reduction effect of the target noise reduction algorithm according to the signal-to-noise ratio of the denoised voice and/or the evaluation information of the user; if the noise reduction effect meets the requirement, sending a push message of the target noise reduction algorithm to a target vehicle; the push message is used for triggering the current noise reduction algorithm of the target vehicle to be switched into the target noise reduction algorithm.
Referring to fig. 1, the target noise reduction algorithm may be located in an offline model library of the host vehicle, in an offline model library of the target vehicle, or from the cloud. Based on this, after S130, S160 and S180, the noise reduction effect of the target noise reduction algorithm is determined according to the signal-to-noise ratio of the denoised speech and/or the evaluation information of the user.
Specifically, calculating the signal-to-noise ratio of the denoised voice, illustratively, if the signal-to-noise ratio exceeds a set threshold, determining that the noise reduction effect of the target noise reduction algorithm is optimal; and if the signal-to-noise ratio does not exceed the set threshold, determining that the noise reduction effect of the target noise reduction algorithm is poor. Evaluation information, such as an evaluation score, input by a user through the in-vehicle terminal is received. And determining whether the noise reduction effect of the target noise reduction algorithm is good or bad according to the evaluation score. And if the noise reduction effect is excellent, the requirement is met, and a push message of the target noise reduction algorithm is sent to the target vehicle. And the target vehicle responds to the push message and switches the current noise reduction algorithm into the target noise reduction algorithm. And if the current noise reduction algorithm of the target vehicle is the target noise reduction algorithm, maintaining the target noise reduction algorithm.
According to the embodiment, the pushing effect of the target noise reduction algorithm is determined, and the target noise reduction algorithm is automatically switched to, so that a user has no perception.
In the above embodiment, the method further comprises: periodically sending a version detection request of an offline model library to the cloud end, so that the cloud end responds to the version detection request and returns a downloading path of a latest version to the vehicle-mounted terminal when detecting that the version of the offline model library is not the latest version; downloading an upgrade package based on the download path; and operating the upgrading package to upgrade the version of the off-line model library to the latest version.
Optionally, the cloud end returns a download path of the latest version to the vehicle when responding to a version detection request of the offline model library sent by the vehicle and detecting that the version of the offline model library is not the latest version; and tracking the task execution progress of the vehicle in real time, and performing corresponding processing when the task execution state is found to be abnormal. For example, the latest version of the download path is re-issued after the interval is set for a certain period of time, or an error is reported to the operation and maintenance personnel.
It should be noted that the above method is a periodic operation, and may be operated while the host vehicle is traveling, or may be operated while the host vehicle is stationary. By periodically upgrading the version of the off-line model library, the voice noise can be reduced directly through the off-line model library under most conditions, and the cost is reduced.
Fig. 2 is a schematic structural diagram of a speech denoising system provided in an embodiment of the present application, including a host vehicle, a target vehicle, and a cloud.
The vehicle is used for collecting noise in the driving process of the vehicle and extracting the characteristic information of the noise; if the target noise reduction algorithm matched with the characteristic information does not exist in the off-line model library of the vehicle, determining the driving environment of the vehicle according to the acquisition information of the sensor and the road condition information; sending the driving environment and the environment in the vehicle to a cloud end; receiving the target noise reduction algorithm from the cloud; denoising the voice of the user of the vehicle by adopting the target denoising algorithm, and adding the target denoising algorithm into the offline model library;
the cloud is used for determining a target vehicle matched with the running environment and the in-vehicle environment of the vehicle, requesting the target noise reduction algorithm from the offline model library of the target vehicle and returning the target noise reduction algorithm to the offline model library of the vehicle;
the driving environment comprises at least one of a POI (point of interest), road condition information, road section information and vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
Optionally, the cloud is configured to perform noise elimination on the characteristic information of the noise to obtain a target noise reduction algorithm; and returning the target noise reduction algorithm to the off-line model base of the vehicle and the target vehicle.
Optionally, the cloud is configured to, in response to a version detection request of an offline model library sent by the host vehicle, return a download path of a latest version to the host vehicle when detecting that the version of the offline model library is not a latest version; and tracking the task execution progress of the vehicle in real time, and performing corresponding processing when the task execution state is found to be abnormal.
Fig. 3 is a schematic structural diagram of another speech denoising system according to an embodiment of the present application, including a host vehicle and at least one candidate vehicle within a set distance range from the host vehicle. In fig. 3, the own vehicle and 4 candidate vehicles travel on 3 lanes in the same direction.
The vehicle is used for collecting noise in the driving process of the vehicle and extracting the characteristic information of the noise; if the target noise reduction algorithm matched with the characteristic information does not exist in the off-line model base of the vehicle, establishing communication connection with the candidate vehicle; acquiring the in-vehicle environment of the candidate vehicles based on the communication connection, and determining a target vehicle matched with the in-vehicle environment and the vehicle from the candidate vehicles; requesting the target noise reduction algorithm from an offline model library of the target vehicle based on a communication connection with the target vehicle; if the target noise reduction algorithm sent by the target vehicle is received, denoising the voice of the user of the vehicle by adopting the target noise reduction algorithm, and adding the target noise reduction algorithm into the offline model library;
the candidate vehicle is used for establishing communication connection with the vehicle and sending the environment in the vehicle to the vehicle based on the communication connection, and the target noise reduction algorithm;
the driving environment comprises at least one of a POI (point of interest), road condition information, road section information and vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
The description of the above speech denoising system can be detailed in the description of the above embodiments, and is not repeated here.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present application and the technical principles employed. It will be understood by those skilled in the art that the present application is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the application. Therefore, although the present application has been described in more detail with reference to the above embodiments, the present application is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present application, and the scope of the present application is determined by the scope of the appended claims.
Claims (9)
1. A voice denoising method is applicable to a vehicle-mounted terminal, and comprises the following steps:
collecting noise in the driving process of the vehicle, and extracting characteristic information of the noise;
if the target noise reduction algorithm matched with the characteristic information does not exist in the offline model library of the vehicle, determining a target vehicle matched with the driving environment and the environment in the vehicle, and requesting the target noise reduction algorithm from the offline model library of the target vehicle;
if a target noise reduction algorithm in the offline model library of the target vehicle is received, denoising the voice of the vehicle user by adopting the target noise reduction algorithm, and adding the target noise reduction algorithm into the offline model library of the vehicle;
the driving environment comprises at least one of a POI (point of interest), road condition information, road section information and vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
2. The method of claim 1, further comprising, after said requesting the target noise reduction algorithm from the off-line model library of the target vehicle:
requesting the target noise reduction algorithm from a cloud if the target noise reduction algorithm does not exist in the offline model library of the target vehicle;
denoising the voice of the vehicle user by adopting the target denoising algorithm, and adding the target denoising algorithm into an offline model library of the vehicle and the target vehicle through the cloud;
the target noise reduction algorithm is stored in a cloud end in advance, or the target noise reduction algorithm is obtained by carrying out noise elimination on the characteristic information of the noise through the cloud end.
3. The method of claim 1, wherein determining a target vehicle matching the driving environment of the host vehicle and the in-vehicle environment and requesting the target noise reduction algorithm from an off-line model library of the target vehicle comprises:
determining the driving environment of the vehicle according to the acquisition information and the road condition information of the sensor;
and sending the running environment and the in-vehicle environment of the vehicle to a cloud end, so that the cloud end can determine a target vehicle matched with the running environment and the in-vehicle environment of the vehicle, request the target noise reduction algorithm from an offline model library of the target vehicle, and return the target noise reduction algorithm to the offline model library of the vehicle.
4. The method of claim 1, wherein determining a target vehicle matching the driving environment of the host vehicle and the in-vehicle environment and requesting the target noise reduction algorithm from an off-line model library of the target vehicle comprises:
searching at least one candidate vehicle in a set distance range with the vehicle, and establishing communication connection with each candidate vehicle;
acquiring the in-vehicle environment of each candidate vehicle based on the communication connection, and determining a target vehicle matched with the in-vehicle environment and the vehicle from each candidate vehicle;
requesting the target noise reduction algorithm from an offline model library of the target vehicle based on a communication connection with the target vehicle.
5. The method of claim 1, further comprising, after said denoising the host-vehicle user's speech using said target noise reduction algorithm:
determining the noise reduction effect of the target noise reduction algorithm according to the signal-to-noise ratio of the denoised voice and/or the evaluation information of the user;
if the noise reduction effect meets the requirement, sending a push message of the target noise reduction algorithm to the target vehicle;
the push message is used for triggering the current noise reduction algorithm of the target vehicle to be switched into the target noise reduction algorithm.
6. A method according to claim 2 or 3, characterized in that the method further comprises:
periodically sending a version detection request of an offline model library to the cloud end, so that the cloud end responds to the version detection request and returns a downloading path of a latest version to the vehicle-mounted terminal when detecting that the version of the offline model library is not the latest version;
downloading an upgrade package based on the download path;
and operating the upgrading package to upgrade the version of the off-line model library to the latest version.
7. A speech denoising system, comprising: the vehicle, the target vehicle and the cloud;
the vehicle is used for collecting noise in the driving process of the vehicle and extracting the characteristic information of the noise; if the target noise reduction algorithm matched with the characteristic information does not exist in the off-line model library of the vehicle, determining the driving environment of the vehicle according to the acquisition information of the sensor and the road condition information; sending the driving environment and the environment in the vehicle to a cloud end; receiving the target noise reduction algorithm from the cloud; denoising the voice of the user of the vehicle by adopting the target denoising algorithm, and adding the target denoising algorithm into the offline model library;
the cloud is used for determining a target vehicle matched with the running environment and the in-vehicle environment of the vehicle, requesting the target noise reduction algorithm from the offline model library of the target vehicle and returning the target noise reduction algorithm to the offline model library of the vehicle;
the driving environment comprises at least one of a POI (point of interest), road condition information, road section information and vehicle speed; the in-vehicle environment includes at least one of engine information, a window state, and an audio-video playback state.
8. The system of claim 7, wherein the cloud is configured to perform noise cancellation on the characteristic information of the noise to obtain a target noise reduction algorithm; and returning the target noise reduction algorithm to the off-line model base of the vehicle and the target vehicle.
9. The system according to claim 7 or 8, wherein the cloud is configured to, in response to a version detection request of an offline model library sent by a host vehicle, return a download path of a latest version to the host vehicle when detecting that a version of the offline model library is not a latest version; and tracking the task execution progress of the vehicle in real time, and performing corresponding processing when the task execution state is found to be abnormal.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011449222.3A CN112735457B (en) | 2020-12-11 | 2020-12-11 | Voice denoising method and system |
| JP2021069324A JP7153759B2 (en) | 2020-12-11 | 2021-04-15 | Audio noise removal method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011449222.3A CN112735457B (en) | 2020-12-11 | 2020-12-11 | Voice denoising method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112735457A CN112735457A (en) | 2021-04-30 |
| CN112735457B true CN112735457B (en) | 2022-05-17 |
Family
ID=75599267
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011449222.3A Active CN112735457B (en) | 2020-12-11 | 2020-12-11 | Voice denoising method and system |
Country Status (2)
| Country | Link |
|---|---|
| JP (1) | JP7153759B2 (en) |
| CN (1) | CN112735457B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113257272A (en) * | 2021-06-29 | 2021-08-13 | 深圳小米通讯技术有限公司 | Voice signal processing method and device, electronic equipment and storage medium |
| CN113628632A (en) * | 2021-07-26 | 2021-11-09 | 北京佳讯飞鸿电气股份有限公司 | Voice noise reduction method, device, equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110197670A (en) * | 2019-06-04 | 2019-09-03 | 大众问问(北京)信息科技有限公司 | Audio defeat method, apparatus and electronic equipment |
| CN110223669A (en) * | 2019-07-01 | 2019-09-10 | 天津职业技术师范大学(中国职业培训指导教师进修中心) | Processing method, the device and system of car noise |
| CN111128192A (en) * | 2019-12-31 | 2020-05-08 | 深圳市优必选科技股份有限公司 | Voice recognition noise reduction method, system, mobile terminal and storage medium |
| CN112053701A (en) * | 2020-09-11 | 2020-12-08 | 北京小米移动软件有限公司 | Sound pickup control method, sound pickup control apparatus, sound pickup control system, sound pickup device, and sound pickup medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107210044B (en) * | 2015-01-20 | 2020-12-15 | 杜比实验室特许公司 | Modeling and reduction of noise in unmanned aerial vehicle propulsion systems |
| US9899018B2 (en) | 2016-06-24 | 2018-02-20 | GM Global Technology Operations LLC | Method, system and apparatus for addressing road noise |
| US10304329B2 (en) * | 2017-06-28 | 2019-05-28 | Zendrive, Inc. | Method and system for determining traffic-related characteristics |
| EP3678385B1 (en) * | 2017-08-30 | 2023-01-04 | Panasonic Intellectual Property Management Co., Ltd. | Sound pickup device, sound pickup method, and program |
| US10347236B1 (en) | 2018-02-28 | 2019-07-09 | Harman International Industries, Incorporated | Method and apparatus for continuously optimized road noise cancellation |
-
2020
- 2020-12-11 CN CN202011449222.3A patent/CN112735457B/en active Active
-
2021
- 2021-04-15 JP JP2021069324A patent/JP7153759B2/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110197670A (en) * | 2019-06-04 | 2019-09-03 | 大众问问(北京)信息科技有限公司 | Audio defeat method, apparatus and electronic equipment |
| CN110223669A (en) * | 2019-07-01 | 2019-09-10 | 天津职业技术师范大学(中国职业培训指导教师进修中心) | Processing method, the device and system of car noise |
| CN111128192A (en) * | 2019-12-31 | 2020-05-08 | 深圳市优必选科技股份有限公司 | Voice recognition noise reduction method, system, mobile terminal and storage medium |
| CN112053701A (en) * | 2020-09-11 | 2020-12-08 | 北京小米移动软件有限公司 | Sound pickup control method, sound pickup control apparatus, sound pickup control system, sound pickup device, and sound pickup medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112735457A (en) | 2021-04-30 |
| JP2022093229A (en) | 2022-06-23 |
| JP7153759B2 (en) | 2022-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9849865B2 (en) | Emergency braking system and method of controlling the same | |
| WO2023201964A1 (en) | Vehicle following target determination method and apparatus, and device and medium | |
| CN112735457B (en) | Voice denoising method and system | |
| US11631326B2 (en) | Information providing system, server, onboard device, vehicle, storage medium, and information providing method | |
| US10817736B2 (en) | System and methods for identifying unoccupied parking positions | |
| US11361555B2 (en) | Road environment monitoring device, road environment monitoring system, and road environment monitoring program | |
| US20200133303A1 (en) | Map information system | |
| US20200104290A1 (en) | Map information system | |
| CN111352142A (en) | Indoor parking positioning method and device, electronic equipment and medium | |
| WO2017015882A1 (en) | Navigation device and navigation method | |
| US20230351823A1 (en) | Information processing device, information processing method and program | |
| US11631328B2 (en) | Determination device, determination method, and non-transitory computer readable storage medium storing determining program | |
| CN113253278B (en) | Parking space identification method and device and computer storage medium | |
| WO2022147785A1 (en) | Autonomous driving scenario identifying method and apparatus | |
| CN114426030B (en) | Pedestrian passing intention estimation method, device, equipment and automobile | |
| JP6303795B2 (en) | Route search system and route search method | |
| CN117476005A (en) | Roof tent control method, system, vehicle and storage medium based on voice recognition | |
| CN114264310A (en) | Positioning and navigation method, device, electronic equipment and computer storage medium | |
| CN115757628A (en) | Vehicle collision detection method, device, equipment and medium | |
| WO2021006110A1 (en) | Vehicle control device | |
| CN111033423A (en) | Method for evaluating a positioning system of an autonomous vehicle | |
| CN111008289A (en) | Image processing apparatus, image processing method, and non-transitory readable recording medium storing program | |
| US20240005672A1 (en) | Information collection system, server, and information collection method | |
| TWI625259B (en) | Method and system for detecting driving situation by mobile device | |
| JP6269677B2 (en) | Driving support system, method and program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |