KR101601523B1 - Method and apparatus for identification of noise sources in automotive using noise fingerprint - Google Patents

Abstract

The present invention relates to an apparatus for recognizing a noise source of a vehicle and a noise source recognizing method using a noise fingerprint. The noise source recognition method of a vehicle, comprises: a step of receiving a query noise signal generated from the vehicle; a step of generating a query noise fingerprint by extracting characteristics of the query noise signal using a spectrogram; and a step of searching database for a noise fingerprint similar to the query noise fingerprint and recognizing a noise source of the query sound signal.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a noise source recognition apparatus for a vehicle and a noise source recognition method using a noise fingerprint,

The present invention relates to a noise source recognition apparatus for a vehicle and a noise source recognition method using a noise fingerprint, in which a database on noise generated in a vehicle is built in advance and a noise signal similar to a fingerprint of a noise currently generated is searched in a database to recognize the cause of noise .

The automobile is made up of many parts, and the noise generated from each part is variously caused by the noise accompanied by the resonance at a specific frequency and the noise generated at the joint part with the other part. As described above, a noise or a noise occurring in an automobile is collectively referred to as BSR (Buzz, Squeak, Rattle) noise.

BSR noise is an important factor that determines the quality satisfaction of the vehicle and it is the most important customer complaint about NVH (Noise, Vibration, Harshness) of the vehicle in recent years.

These BSR noises are divided into Buzz, which is caused by parts of the panel itself falling like a drum, a squeak caused by friction between parts, and a rattle caused by a collision between parts. BSR noise varies depending on road surface condition, driving condition, body structure and interior parts characteristics, vehicle type and vehicle type, noise level, and location of noise source. Therefore, The subjective evaluation that you feel is different.

According to the importance of BSR noise problem, domestic and foreign automobile manufacturers use various approaches in their own way. However, it is common to use the so-called 'Find and Solve' method, which relies on subjective evaluation of relevant experts in the pre- Fix) method.

However, such a method has difficulty in quantitative and objective BSR evaluation by reflecting diversity of BSR noise source, complexity of occurrence mechanism, and subjectivity of evaluation. In addition, if a customer complains about the BSR noise after the death, it is determined whether the parts are repaired according to the subjective evaluation of the repair shop engineer in order to judge the failure, thereby increasing friction and complaints between the automobile manufacturer and the consumer .

KR 1020010009050E KR 1020050011919 A KR 1019950013806 A KR 1020100035516 A KR 1020070069837 A

SUMMARY OF THE INVENTION The present invention has been conceived in order to solve the problems of the prior art described above, and it is an object of the present invention to provide a database for noise generated in a vehicle and to search a database for a noise signal similar to a fingerprint of a current noise, And a noise source recognition method using a noise fingerprint.

According to an aspect of the present invention, there is provided a noise source recognition method for a vehicle, comprising: receiving a query noise signal generated in a vehicle; extracting characteristics of the query noise signal using a spectrogram; Generating a noise fingerprint; searching a database for a noise fingerprint similar to the query noise fingerprint; and recognizing the noise source of the query noise signal based on the search result.

In addition, the step of generating the query silence fingerprint may include a preprocessing step of resampling the query noise signal so that the sampling rate of the query noise signal coincides with the sampling rate of the reference noise signal in the database, performing spectrogram analysis on the query noise signal And extracting the feature through the spectrogram analysis.

The spectrogram analysis step may include calculating a spectrum for each time frame with respect to the query noise signal.

The feature extraction step divides the frequency axis of the spectrogram into a predetermined number of critical frequency bands, calculates energy of each time frame and a critical frequency band, and binarizes the spectrogram by comparing energy sizes with adjacent bands, And a matrix is calculated.

The database searching step may include deriving a matching matrix of the query noise fingerprint and the reference noise fingerprint in the database and calculating a matching rate of the query noise fingerprint and the reference noise fingerprint using the matching matrix .

The noise source recognition apparatus for a vehicle according to an embodiment of the present invention includes a noise input unit for receiving a query noise signal generated in a vehicle, a noise fingerprint generation unit for generating a query noise fingerprint by extracting characteristics of the query noise signal, And a matching unit for searching a database for a noise fingerprint similar to the query noise fingerprint generated by the noise fingerprint generation unit and recognizing the noise source of the query noise signal.

The noise input unit may include a microphone for measuring the sound pressure of the query noise signal, and an accelerometer for measuring vibration of the query noise signal.

Meanwhile, a method for generating a noise fingerprint of a vehicle noise recognition device according to an embodiment of the present invention includes steps of measuring a noise signal generated in a vehicle, performing a spectrogram analysis on a noise signal measured in the vehicle, Extracting features of the noise signal through the spectrogram analysis, and storing the characteristics of the noise signal in a database as noise fingerprints.

The spectrogram analysis step may include calculating a spectrum for each time frame with respect to the query noise signal.

The feature extraction step divides the frequency axis of the spectrogram into a predetermined number of critical frequency bands, calculates energy of each time frame and a critical frequency band, and binarizes the spectrogram by comparing energy sizes with adjacent bands, And a matrix is calculated.

The present invention preliminarily builds a database of noise generated in a vehicle and searches the database for noise signals similar to the fingerprints of the current noise to recognize the cause of the noise. Therefore, the present invention can objectively and accurately identify the cause of the noise generated in the vehicle and promptly provide it.

In addition, according to the present invention, when a noise of a new pattern that is not present in an existing database is generated, the noise source recognition performance and usability can be increased by updating it in the database.

In addition, since the present invention uses not only a microphone for measuring noise but also an accelerometer for measuring vibration, it is possible to detect a noise source without being affected by the signal-to-noise ratio even in a high background noise level environment.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a noise source recognition apparatus for a vehicle according to an embodiment of the present invention; FIG.
FIG. 2 is a block diagram of the noise silencing unit shown in FIG. 1. FIG.
3 is a diagram illustrating an example of a noise fingerprint generation process according to the present invention.
FIG. 4 is an exemplary diagram illustrating a matching matrix derivation process according to the present invention; FIG.
5 is a flowchart illustrating a noise source recognition method of a vehicle according to an embodiment of the present invention.
FIG. 6 is a flowchart illustrating a noise fingerprint generation process in FIG. 5; FIG.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

The present invention relates to a method and apparatus for measuring the generation and generation mechanism of a BSR noise source occurring in a current vehicle by comparing the noise and vibration of a vehicle in which BSR (Buzz, Squeak, Rattle) . The present invention extracts features of BSR noise generated in a vehicle in advance and registers the extracted noise features in a database using an acoustic fingerprint in order to identify the location and generation mechanism of the noise source.

FIG. 1 is a block diagram illustrating a noise source recognition apparatus for a vehicle according to an embodiment of the present invention. FIG. 2 is a block diagram of a silence fingerprint generation unit shown in FIG. 1. FIG. FIG. 4 is a diagram illustrating a process of deriving a matching matrix according to an embodiment of the present invention. Referring to FIG.

1, a noise source recognition apparatus for a vehicle includes a noise input unit 10, a noise fingerprint generation unit 20, a matching unit 30, a database (DB) 40, (50).

The noise input unit 10 measures a query noise signal generated in the vehicle for a predetermined time. Here, the query noise signal means BSR (Buzz, Squeak, Rattle) noise which may be the cause of the problem but may be a cause of occurrence.

The noise input unit 10 includes a microphone for measuring a sound pressure of a noise signal and an accelerometer for measuring vibrations, and is installed on a door hinge, a brake drum, a seat, or the like.

The noise fingerprint generation unit 20 generates a sound fingerprint (hereinafter referred to as a query noise fingerprint) for the query noise signal input through the noise input unit 10. [ The noise fingerprint generation unit 20 includes a preprocessing unit 21, an analysis unit 22, and a feature extraction unit 23 as shown in FIG.

The preprocessing unit 21 confirms whether the sampling rate of the query noise signal is equal to the sampling rate of the reference noise signal stored in advance in the DB 40 (for example, 51.2k [sample / sec]). The preprocessing unit 21 resamplifies the sampling rate of the query noise signal and the sampling rate of the reference noise signal to coincide with the sampling rate of the reference noise signal.

The analyzer 22 analyzes the query noise signal using a spectrogram. In other words, the analyzer 22 obtains the spectrogram to observe the frequency component of the query noise signal which changes with time. In this embodiment, spectrum analysis using Fourier transform is not suitable because BSR noise is a transient signal that occurs during a short period of time and disappears. Therefore, a spectrogram using a local Fourier transform is used, .

The feature extraction unit 23 extracts features from the query noise signal. At this time, the feature extraction unit 23 divides the frequency axis of the spectrogram output from the analysis unit 22 into a predetermined number (for example, 24) of critical frequency bands, and calculates energy corresponding to each time frame and band. This is to speed up the comparison operation with data in the DB 40 by reducing the dimension of the spectrogram, and to extract the feature of the query noise signal by modeling the human voice recognition characteristic.

The feature extracting unit 23 compares energy magnitudes with adjacent frequency axis bands using time-frame-by-band energy, and binarizes them into binary values of 0 or 1. This binarization process can be expressed as [1].

Where n is a time base frame, m is a threshold frequency band divided by a predetermined number (e.g., 24), P (n, m) is the power of n frame m bands, F (n, m) to be. For example, F (n, m) becomes 1 if the upper band energy is greater than the energy of the current band in the same frame, and F (n, m) has a value of 0 in the opposite case.

The feature extraction unit 23 extracts characteristics of a query noise signal through a binarization process. This feature is extracted in the form of a feature matrix, as shown in FIG. 3, as a voice fingerprint (a query noise fingerprint) of a query noise signal. The dimension of the row of the feature sequence coincides with the time-frame axis and the dimension of the column decreases by one from 24, the total number of critical frequency bands, to 23.

The matching unit 30 searches the DB 40 for a reference noise fingerprint similar to the query noise fingerprint outputted from the noise fingerprint generating unit 20 and recognizes the noise source of the query noise fingerprint. At this time, the matching unit 30 uses the query noise fingerprint as index information for data retrieval in the DB 40 constructed in advance.

The matching unit 30 compares the feature matrix Q of the query noise signal with the feature matrix D of the DB 40 and calculates the similarity of the two matrices. To this end, the matching unit 30 calculates a matching matrix of the feature matrix Q of the query noise signal and the feature matrix D of the DB 40 using [Formula 2].

The element of the matching matrix R has a value of 0 when the characteristic matrix Q of the query noise signal is matched with a corresponding element of the characteristic matrix D of the DB 40 and has a value of 1 when mismatching .

와 쿼리 소음 신호의 특징 행렬(Q)의 행의 크기

는 다르다. 따라서, 두 행렬 중에서 열의 크기가 작은 차원으로 일치시켜, 시간 프레임 축으로 이동시켜가면서 두 행렬을 비교한다.For example, as shown in FIG. 4, the dimensions of the columns of two feature matrices are always the same, but the dimensions of the rows of the two matrix are not equal because the number of time frames is different. That is, the size of the row of the feature matrix D of the DB 40

And the size of the row of the feature matrix Q of the query noise signal

Is different. Therefore, two columns are matched with a small dimension of the column, and two matrices are compared while moving to the time frame axis.

The matching unit 30 determines the matching ratio (degree of similarity) by how many 0's are contained in the elements of the matching matrix R (?). Here, the matching ratio is represented by a score between 0 and 100. If the query noise signal coincides with the data of the DB 40, the matching matrix R becomes a zero matrix and the matching ratio becomes 100%.

The matching rate can be expressed as [Numerical 3].

The DB 40 is constructed in advance as information on noise that can be generated in the vehicle. The noise information in the DB 40 is information on reference noise signals used for noise source identification of the query noise signal. Such reference noise signal related information includes reference noise fingerprint and cause of occurrence, countermeasure, occurrence position, measurement condition, and the like. Here, the measurement conditions include information such as measurement date, vehicle type, sensor type, measurement temperature, measurement condition (traveling, door opening, etc.).

The data stored in the DB 40 can be updated in the development stage of the vehicle or in the post-A / S process.

The output unit 50 visualizes and / or audibly displays the search result output from the matching unit 30 and outputs the result. The output unit 50 may be implemented as a liquid crystal display, a transparent display, a light emitting diode (LED) display, a flexible display, or the like.

FIG. 5 is a flowchart illustrating a noise source recognition method of a vehicle according to an embodiment of the present invention, and FIG. 6 is a flowchart illustrating a process of generating a noise fingerprint in FIG.

First, the noise source recognition apparatus of the vehicle receives the query noise signal through the noise input unit 10 (S11). At this time, the user executes the noise source recognition function and inputs a query noise signal to the specific period of the noise signal measured in advance along with the search condition on the execution screen through the input means (not shown). Here, the search conditions include a car model, a sensor type, a measurement situation (traveling, door opening, braking, etc.), and a measurement temperature.

The noise fingerprint generation unit 20 of the noise source recognition apparatus of the vehicle generates a query noise fingerprint for the query noise signal (S12).

Hereinafter, the process of generating a query noise fingerprint will be described in more detail.

First, the preprocessing unit 21 of the noise fingerprint generation unit 20 preprocesses the query noise signal inputted through the noise input unit 10 (S121). The preprocessing unit 21 resamples the query noise signal so that the sampling rate of the query noise signal coincides with the sampling rate of the reference noise signal in the DB 40.

The noise fingerprint generation unit 20 performs spectrogram analysis on the query noise signal passed through the preprocessing unit 21 (S122). That is, the noise fingerprint generation unit 20 obtains a spectrogram for the query noise signal.

The feature extraction unit 23 extracts the feature matrix of the query noise signal from the spectrogram (S123). The feature extraction unit 23 divides the frequency axis of the spectrogram into predetermined threshold frequency bands and calculates energy for each frame and band. Then, the feature extraction unit 23 calculates the feature matrix by binarizing the spectrogram through energy comparison with adjacent bands in the band axis (frequency axis). The feature matrix represents the voice fingerprint of the query noise signal.

The matching unit 30 searches the database 40 for a noise fingerprint similar to the query noise fingerprint generated by the noise fingerprint generating unit 20 (S13). In other words, the matching unit 30 compares the feature matrix of the query noise signal with the feature matrix of the reference noise stored in the DB 40 to calculate a matching matrix. Then, the matching unit 30 calculates the matching ratio (similarity) based on the matching matrix.

At this time, the matching unit 30 can detect the reference noise signal having the noise fingerprint whose matching rate with the query noise fingerprint is not less than the reference (for example, 70%). Alternatively, the matching unit 30 may detect a certain number of reference noise signals in descending order of the matching ratio.

The matching unit 30 recognizes the noise source of the query noise signal based on the search result (S14). For example, if the matching unit 30 searches the DB 40 for a reference noise fingerprint that is 100% identical to the query noise fingerprint, it outputs the cause of the reference noise signal corresponding to the reference noise fingerprint, .

When a plurality of results are retrieved, the matching unit 30 displays the retrieval result on the output unit 50. If any one of the retrieval results is selected, the matching unit 30 reads the detailed contents of the retrieval result from the DB 40 and displays do. Accordingly, the user can recognize the noise source of the query noise signal through the detailed contents browsing. Here, the details include the cause and location of the query noise signal, countermeasures, and notes.

In addition, if the noise fingerprint similar to the sound fingerprint of the query noise signal is not retrieved from the DB 40, the noise source recognition apparatus of the present invention adds the noise fingerprint of the query noise signal and related information related thereto to the DB 40 .

While the present invention has been described in connection with what is presently considered to be the most practical and preferred embodiments, it is to be understood that the invention is not limited to the disclosed embodiments. That is, within the scope of the present invention, all of the components may be selectively coupled to one or more of them. In addition, although all of the components may be implemented as one independent hardware, some or all of the components may be selectively combined to perform a part or all of the functions in one or a plurality of hardware. As shown in FIG. The codes and code segments constituting the computer program may be easily deduced by those skilled in the art. Such a computer program can be stored in a computer-readable storage medium, readable and executed by a computer, thereby realizing an embodiment of the present invention. As the storage medium of the computer program, a magnetic recording medium, an optical recording medium, a carrier wave medium, or the like may be included.

It is to be understood that the above-described "comprises " means that the constituent element can be implanted unless specifically stated to the contrary, it is to be understood that other constituent elements may be included, Should be interpreted. All terms, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Commonly used terms, such as predefined terms, should be interpreted to be consistent with the contextual meanings of the related art, and are not to be construed as ideal or overly formal, unless expressly defined to the contrary.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

10: Noise input section
20: Noise fingerprint generator
30:
40: DB
50: Output section

Claims (10)

Receiving a query noise signal generated in a vehicle,
Generating a query noise fingerprint by extracting features of the query noise signal using a spectrogram;
Searching a database for a noise fingerprint similar to the query noise fingerprint,
And recognizing a noise source of the query noise signal based on a result of the search. The method according to claim 1,
Wherein the query noise fingerprint generation step comprises:
A preprocessing step of re-sampling the sampling noise signal so that a sampling rate of the query noise signal coincides with a sampling rate of the reference noise signal in the database;
Performing a spectrogram analysis on the query noise signal;
And extracting features through the spectrogram analysis. ≪ RTI ID = 0.0 > 11. < / RTI > 3. The method of claim 2,
Wherein the spectrogram analysis step comprises:
And a spectrum for each of the time frames is calculated for the query noise signal. 3. The method of claim 2,
The feature extraction step may include:
Wherein the frequency matrix of the spectrogram is divided into a predetermined number of threshold frequency bands, energy of each time frame and a threshold frequency band is calculated, and the spectrogram is binarized by comparing energy sizes with adjacent bands to calculate a feature matrix. A method for recognizing a noise source of a vehicle. The method according to claim 1,
Wherein the database searching step comprises:
Deriving a matching matrix of the query noise fingerprint and the reference noise fingerprint in the database;
And calculating a matching rate of the query noise fingerprint and the reference noise fingerprint using the matching matrix. A noise input unit for receiving a query noise signal generated in the vehicle;
A noise fingerprint generation unit for extracting characteristics of the query noise signal to generate a query noise fingerprint,
And a matching unit for searching the database for a noise fingerprint similar to the query noise fingerprint generated by the noise fingerprint generation unit and recognizing a noise source of the query noise signal. The method according to claim 6,
The noise input unit may include:
A microphone for measuring the sound pressure of the query noise signal;
And an accelerometer for measuring the vibration of the query noise signal. Measuring a noise signal generated in the vehicle,
Performing a spectrogram analysis on a noise signal measured in the vehicle;
Extracting features of the noise signal through the spectrogram analysis;
And storing the characteristic of the noise signal as a noise fingerprint in a database. 9. The method of claim 8,
Wherein the spectrogram analysis step comprises:
And a spectrum for each time frame is calculated for the query noise signal. 9. The method of claim 8,
The feature extraction step may include:
Wherein the frequency matrix of the spectrogram is divided into a predetermined number of threshold frequency bands, energy of each time frame and a threshold frequency band is calculated, and the spectrogram is binarized by comparing energy sizes with adjacent bands to calculate a feature matrix. A method for generating a noise fingerprint of a vehicle noise recognition device.

Images