CN117457026A - Noise detection system and method for riding equipment - Google Patents
Noise detection system and method for riding equipment Download PDFInfo
- Publication number
- CN117457026A CN117457026A CN202311152024.4A CN202311152024A CN117457026A CN 117457026 A CN117457026 A CN 117457026A CN 202311152024 A CN202311152024 A CN 202311152024A CN 117457026 A CN117457026 A CN 117457026A
- Authority
- CN
- China
- Prior art keywords
- abnormal
- sound
- noise
- noise detection
- detection
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 209
- 238000000034 method Methods 0.000 title claims abstract description 26
- 230000002159 abnormal effect Effects 0.000 claims abstract description 198
- 238000013145 classification model Methods 0.000 claims abstract description 10
- 230000008569 process Effects 0.000 claims abstract description 7
- 238000000605 extraction Methods 0.000 claims description 32
- 239000000284 extract Substances 0.000 claims description 12
- 238000012797 qualification Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000003993 interaction Effects 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000010183 spectrum analysis Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 210000005069 ears Anatomy 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H11/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by detecting changes in electric or magnetic properties
- G01H11/06—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by detecting changes in electric or magnetic properties by electric means
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention provides a noise detection system and method of riding equipment, which relates to the technical field of riding equipment detection, wherein a flywheel of the riding equipment and/or a seat sleeve are/is set as detection points, and the noise detection system comprises the following components: collecting the sound of a detection point of riding equipment in the riding detection process to obtain audio data; extracting audio characteristics of the audio data, and carrying out noise detection according to the audio characteristics to obtain a noise detection result; and extracting an abnormal section from the audio data according to the audio characteristics, carrying out sound classification on the abnormal section, carrying out abnormal sound judgment according to the classification result to obtain an abnormal sound judgment result, and prompting a detector that riding equipment detection does not pass when the noise detection result indicates noise or the abnormal sound judgment result indicates abnormal sound. The method has the beneficial effects that interference noise in the environment is removed through the classification model, the quality of riding equipment is detected through two aspects of noise detection and abnormal sound detection, and the accuracy and the automation and the unified detection standard are realized.
Description
Technical Field
The invention relates to the technical field of riding equipment detection, in particular to a noise detection system and method of riding equipment.
Background
In the prior art, when detecting noise of riding equipment, the following method is often adopted:
the human ear listens: this is the most primitive and closest to the actual detection means, but also has problems 1 such as subjectivity, fatigue, and erroneous judgment rate.
Sound level meter: this is an electronic instrument that can measure sound pressure levels but cannot distinguish between different types and sources of noise.
Spectral analysis: this is a method of frequency decomposing an acoustic signal with a signal processor, which finds out the noise in certain specific frequency bands, but is less effective for electroacoustic products or low energy noise.
Distortion analysis: this is a method of detecting noise by comparing the energy difference of normal and imperfection sounds, but is less accurate for the case of small abnormal or overall sound variations.
Pure tone measurement: this is a method for detecting noise using a sliding frequency signal and slope analysis, which can exhaust electroacoustic noise and mechanical noise on a net, but requires special instruments and software.
From the above method, it can be seen that the disadvantages in the prior art mainly include the following points:
when the noise of the riding equipment is detected manually, simple tools such as a human ear sound meter or a sound level meter are usually adopted, the methods are high in labor cost, the detection standards are not uniform, and the method is easily influenced by subjective judgment of individuals and interference of external environments, so that the noise detection of the riding equipment cannot be unified, accurate and automatic.
When detecting noise of riding equipment, complex tools such as spectrum analysis and distortion analysis are usually adopted, and although the methods can improve detection efficiency and accuracy, the methods have the problems of narrow application range, high cost, requirement of special instruments and software, low test speed and the like.
Disclosure of Invention
In view of the problems existing in the prior art, the present invention provides a noise detection system of a riding device, where a flywheel of the riding device and/or a seat sleeve are set as detection points, including:
the sound acquisition device is used for acquiring the sound of the detection point of the riding equipment in real time in the riding detection process to obtain audio data;
the noise detection module is connected with the sound acquisition device and is used for extracting the audio characteristics of the audio data and carrying out noise detection according to the audio characteristics to obtain a noise detection result;
the abnormal sound detection module is connected with the noise detection module and is used for extracting abnormal sections from the audio data according to the audio characteristics, classifying the abnormal sections in sound and judging abnormal sound according to the classification result to obtain an abnormal sound judgment result;
and the qualification judging module is respectively connected with the noise detecting module and the abnormal sound detecting module and is used for prompting a detector that the riding equipment is not detected when the noise detecting result indicates noise or the abnormal sound judging result indicates abnormal sound.
Preferably, the noise detection module includes:
the characteristic extraction unit is used for respectively calculating the A weight sound pressure level of each data frame in the audio data as the corresponding audio characteristic;
and the noise judgment unit is connected with the characteristic extraction unit and is used for not passing the noise detection as the noise detection result when the A weight sound pressure level exceeds a preset maximum sound pressure level and the duration exceeds the longest duration, and passing the noise detection as the noise detection result when the A weight sound pressure level does not exceed the preset maximum sound pressure level or the duration does not exceed the longest duration.
Preferably, the abnormal sound detection module includes:
an extracting unit, configured to extract, as one of the anomaly segments, each of the continuous data frames in the audio data, where the audio feature of the data frame exceeds a preset anomaly value;
the classification unit is connected with the extraction unit and is used for extracting the characteristics of each abnormal section to obtain corresponding mel cepstrum coefficients, inputting each mel cepstrum coefficient into a pre-trained classification model to obtain a sound classification result of each abnormal section, and eliminating the abnormal sections except for the riding equipment according to the sound classification result;
and the abnormal sound judging unit is connected with the classifying unit and is used for not passing abnormal sound detection as an abnormal sound detection result when the number of the rest abnormal sections is judged to be larger than an abnormal sound number threshold value and the time interval between every two adjacent abnormal sections is in a preset time range, and passing abnormal sound detection as the abnormal sound detection result when the number of the abnormal sections is judged to be not larger than an abnormal sound number threshold value or the time interval between every two adjacent abnormal sections is judged to be not in the preset time range.
Preferably, the extraction unit includes:
a first extraction subunit, configured to mark, in the audio data, the data frame that exceeds a preset noise threshold and that is not higher than the noise threshold as a recording start point;
a second extraction subunit, connected to the first extraction subunit, configured to mark, as a recording end point, a data frame having a first audio feature lower than a preset stop value when the audio feature of the data frame after the recording start point is higher than the abnormal value;
and a third extraction subunit, connected to the first extraction subunit and the second extraction subunit, configured to extract, as one of the abnormal segments, each of the data frames between the recording start point and the recording end point, and a plurality of the data frames before the recording start point and a plurality of the data frames after the recording end point.
Preferably, the abnormal sound detection module further includes a preprocessing unit connected to the extracting unit and the classifying unit, respectively, and configured to perform normalized data processing on the extracted abnormal segment as the abnormal segment.
Preferably, the system further comprises a man-machine interaction interface module, which is respectively connected with the sound collection module, the noise detection module, the abnormal sound detection module and the qualification judgment module, and is used for displaying a waveform chart, a spectrogram, an energy chart and the audio characteristics corresponding to the audio data, and the noise detection result and the abnormal sound judgment result.
The invention also provides a noise detection method of the riding equipment, which is applied to the noise detection system and comprises the following steps:
step S1, the noise detection system collects the sound of a detection point of the riding equipment in real time in the riding detection process to obtain audio data;
step S2, the noise detection system extracts the audio characteristics of the audio data, and performs noise detection according to the audio characteristics to obtain a noise detection result;
step S3, the noise detection system extracts abnormal sections from the audio data according to the audio characteristics, carries out sound classification on the abnormal sections, and carries out abnormal sound judgment according to classification results to obtain abnormal sound judgment results;
step S4, the noise detection system judges whether the noise detection result represents noise or whether the abnormal sound judgment result represents abnormal sound or not:
if yes, prompting a detector that the riding equipment is not passed;
if not, prompting the detection personnel that the riding equipment passes the detection.
Preferably, the step S2 includes:
step S21, the noise detection system calculates A weight sound pressure level of each data frame in the audio data as the corresponding audio feature;
step S22, the noise detection system determines whether the a weight sound pressure level exceeds a preset maximum sound pressure level and the duration exceeds a maximum duration:
if yes, the noise detection is not passed as the noise detection result;
if not, the noise detection is passed as the noise detection result.
Preferably, the step S3 includes:
step S31, the noise detection system respectively extracts each continuous data frame with the audio characteristic exceeding a preset abnormal value in the audio data as one abnormal section;
step S32, the noise detection system performs feature extraction on each abnormal section to obtain a corresponding Mel cepstrum coefficient, respectively inputs each Mel cepstrum coefficient into a pre-trained classification model to obtain a sound classification result of each abnormal section, and eliminates the abnormal section except for the riding equipment according to the sound classification result;
step S33, the noise detection system determines whether the number of remaining abnormal segments is greater than an abnormal sound number threshold and whether a time interval between adjacent abnormal segments is within a preset time range:
if yes, the abnormal sound detection is not passed as the abnormal sound detection result;
if not, the abnormal sound detection is used as the abnormal sound detection result.
Preferably, the step S31 includes:
step S311, the noise detection system marks, as a recording start point, the data frame that exceeds a preset noise threshold and that is not higher than the noise threshold in the audio data;
step S312, the noise detection system marks the data frame with the first audio feature lower than a preset stopping value as a recording end point when the audio feature of the data frame after the recording start point is higher than the abnormal value;
in step S313, the noise detection system extracts each of the data frames between the recording start point and the recording end point, and the plurality of data frames before the recording start point and the plurality of data frames after the recording end point as one of the abnormal segments.
The technical scheme has the following advantages or beneficial effects:
1) When abnormal sound detection is carried out, the interference noise in the environment is removed through the classification model, so that the influence of the interference noise in the environment on noise detection is reduced, and the detection mode is more accurate than that of a traditional sound level meter;
2) The noise detection system is adopted to collect the audio, and the problem that the detection standard is not uniform due to the subjective judgment of manpower and the influence of external interference in the traditional manual detection mode can be avoided by analyzing the audio, so that the noise detection of the riding equipment cannot be unified, accurate and automatic;
3) Common and commonly used equipment is adopted in the noise detection system, professional instruments and software are not needed, the problems of narrow application range, high cost and the like in the traditional mode can be avoided, the detection speed is higher, and the labor cost can be saved compared with the manual detection mode.
Drawings
FIG. 1 is a schematic diagram of a noise detection system of a riding device according to a preferred embodiment of the present invention;
FIG. 2 is a flow chart of a method for detecting noise of a riding device according to a preferred embodiment of the present invention;
FIG. 3 is a schematic flow chart of step S2 in the preferred embodiment of the present invention;
FIG. 4 is a schematic flow chart of step S3 according to the preferred embodiment of the present invention;
fig. 5 is a schematic flow chart of step S31 in the preferred embodiment of the invention.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The present invention is not limited to the embodiment, and other embodiments may fall within the scope of the present invention as long as they conform to the gist of the present invention.
In accordance with the foregoing and other objects and features of the present invention, there is provided a noise detection system for a riding device, a flywheel of the riding device, and/or a seat sleeve as detection points, as shown in fig. 1, comprising:
the sound collection device 1 is used for collecting the sound of a detection point of the riding equipment in real time in the riding detection process to obtain audio data;
the noise detection module 2 is connected with the sound acquisition device 1 and is used for extracting the audio characteristics of the audio data and carrying out noise detection according to the audio characteristics to obtain a noise detection result;
the abnormal sound detection module 3 is connected with the noise detection module 2 and is used for extracting abnormal sections from the audio data according to the audio characteristics, carrying out sound classification on the abnormal sections, and carrying out abnormal sound judgment according to the classification result to obtain an abnormal sound judgment result;
and the qualification judging module 4 is respectively connected with the noise detecting module and the abnormal sound detecting module and is used for prompting the detecting personnel that the riding equipment detection does not pass when the noise detecting result indicates that the noise exists or the abnormal sound judging result indicates that the abnormal sound exists.
Specifically, in this embodiment, noise detection of the riding device is divided into two parts: noise and abnormal sound, the noise mainly aims at a certain sound which occurs for a long time, and the abnormal sound mainly aims at a periodically occurring sound;
the sound collection device 1 mainly includes: the two directional microphones are preferably set towards two detection points on the riding equipment respectively by adopting a mountain-shaped C02, and then are connected with a sound card, in the embodiment, a sound card UMC202HD is preferably adopted, the sound card is connected with an edge diagnosis unit arranged on the computer equipment, a code scanning gun is also connected with the computer for acquiring the product number of the riding equipment to be detected currently, in order to check whether the noise detection system in the embodiment works normally or not, the sound card is connected with the sound card through the earphone for monitoring, if the noise appears in the audio data acquired in real time, but the noise detection system does not prompt that the detection does not pass, the noise detection system does not work normally, the noise detection system in the embodiment adopts common and commonly used equipment, professional instruments and software are not needed, and the problems of narrow application range, high cost and the like in the traditional mode can be avoided;
the abnormal sound detection system in the embodiment is divided into two sound channels to collect real-time audio data of two detection points, so that audio characteristics of the audio data can be extracted, and noise detection and abnormal sound detection are carried out according to the audio characteristics; when the abnormal sound is detected, the interference sound in the environment can be identified through the classification model, other interference sounds are separated and then deleted, only the audio data collected by the riding equipment are reserved, and the interference of the environment to the abnormal sound detection is reduced.
In a preferred embodiment of the present invention, as shown in fig. 1, the noise detection module 2 includes:
a feature extraction unit 21 for calculating a-weighted sound pressure levels of the respective data frames in the audio data as corresponding audio features, respectively;
a noise judgment unit 22 connected to the feature extraction unit 21 for passing no noise detection as a noise detection result when the a-weighted sound pressure level exceeds a preset maximum sound pressure level and the duration exceeds a maximum duration, and passing noise detection as a noise detection result when the a-weighted sound pressure level does not exceed the preset maximum sound pressure level or the duration does not exceed the maximum duration.
Specifically, in this embodiment, first, frequency domain analysis is performed on each data frame in the collected audio data through frequency domain analysis to obtain a corresponding a weight sound pressure level as an audio feature, where the a weight sound pressure level (a-weighted sound pressure level) is a sound pressure level obtained by weighting sound according to auditory characteristics of a human ear. The human ear has different sensitivity to sounds of different frequencies, in particular relatively insensitive to low frequency sounds and relatively sensitive to medium and high frequency sounds. In order to better reflect the auditory perception of the human ear, the international organization for standardization (ISO) introduced an a-weighting curve according to which sounds were weighted to obtain a-weighting sound pressure levels. The A weight sound pressure level is generally expressed in units of decibels (dB), commonly referred to as dB (A). The influence of the intensity of sound on the human ear can be more accurately evaluated using the a-meter weights when measuring the sound. By weighting the sound pressure levels according to the weighting a, the contribution of the low frequency sound is reduced, while the contribution of the medium and high frequency sound is amplified, more in line with the auditory perception of the human ear. The A weight sound pressure level is widely applied to the fields of noise control, environmental noise monitoring and the like and is used for evaluating the influence of sound on human health and environment.
And then, when the sound pressure of the audio data collected in real time exceeds the preset maximum sound pressure level, starting to record the duration, and if the duration is longer than the preset maximum duration, indicating that the noise detection of the riding equipment is not passed, namely, continuous noise can be generated during riding.
In a preferred embodiment of the present invention, as shown in fig. 1, the abnormal sound detection module 3 includes:
an extracting unit 31, configured to extract, as an abnormal segment, each continuous data frame in which an audio feature in the audio data exceeds a preset abnormal value;
the classifying unit 32 is connected with the extracting unit 31 and is used for extracting the characteristics of each abnormal segment to obtain corresponding mel cepstrum coefficients, inputting each mel cepstrum coefficient into a pre-trained classifying model to obtain the sound classifying result of each abnormal segment, and eliminating the abnormal segments except for riding equipment according to the sound classifying result;
the abnormal sound judging unit 33 is connected to the classifying unit 32, and is configured to, when it is judged that the number of remaining abnormal segments is greater than the abnormal sound number threshold and the time interval between adjacent abnormal segments is within the preset time range, not pass abnormal sound detection as an abnormal sound detection result, and when it is judged that the number of abnormal segments is not greater than the abnormal sound number threshold or the time interval between adjacent abnormal segments is not within the preset time range, pass abnormal sound detection as an abnormal sound detection result.
Specifically, in this embodiment, the classification model adopts a voiceprint model, and each of the differences Chang Duan is input into a pre-trained main stream voiceprint model (ECAPA-TDNN) to perform multi-label prediction. When the model is trained, on-site recorded speech sounds, electric drill sounds, machining sounds, footstep sounds, noise and the like which are common in environments such as idle tuning and various abnormal sounds of a body-building chair are used, the trained model can rapidly identify and classify various abnormal sections, when the model predicts, whether each type of sound exists or not can be obtained according to corresponding confidence coefficients (0.0-1.0), if the confidence coefficients are higher than 0.5, the environment interference sounds exist in the abnormal sections are judged, and the corresponding abnormal sections are removed. If all confidence degrees of the prediction results corresponding to the field abnormal section are smaller than 0.5, the abnormal section marked as abnormal sound of one riding device is reserved.
In a preferred embodiment of the present invention, as shown in fig. 1, the extracting unit 31 includes:
a first extraction subunit 311, configured to mark, in the audio data, a data frame that exceeds a preset noise threshold and that is not higher than the noise threshold as a recording start point;
a second extraction subunit 312, connected to the first extraction subunit 311, configured to mark a data frame with a first audio feature lower than a preset suspension value as a recording end point when an audio feature of the data frame after the recording start point is higher than an abnormal value;
a third extraction subunit 313 connected to the first extraction subunit 311 and the second extraction subunit 312, and configured to extract each data frame between the recording start point and the recording end point, and a plurality of data frames before the recording start point and a plurality of data frames after the recording end point as one abnormal segment.
Specifically, in this embodiment, noise detection is performed on a riding device, firstly, an abnormal segment in audio data is extracted, and an abnormal segment extractor is used for extracting the abnormal segment, and the main extraction process is as follows:
firstly, a noise threshold, an abnormal value and an abort value are calibrated, (the noise threshold is higher than the noise threshold and represents that the sound emitted by the riding equipment can be heard by human ears and is considered as noise, the abnormal value is higher than the abnormal value and represents that the sound emitted by the riding equipment is not abnormal sound which should appear when the riding equipment is normal, the abnormal sound is lower than the abort value and represents that the abnormal sound stops), audio data are collected in real time and are expressed in the form of a data stream, when the data frame which exceeds the preset noise threshold and is not higher than the noise threshold is appeared, the data frame is taken as a recording starting point to pay attention to the subsequent data stream, if the sound pressure value is continuously increased and exceeds the abnormal value, the abnormal sound is generated, the data frame which is lower than the abort value is taken as a recording ending point of the abnormal sound, then the data frame between the recording starting point and the recording ending point is taken as an abnormal section (namely, the abnormal sound is once), and in order to ensure accuracy, a plurality of data frames in a small section of audio (generally ranging from 10ms to 30 ms) and a small section of audio (generally ranging from 10ms to 30 ms) are also selected from the small section of audio frame which is recorded.
In this way, all abnormal segments in the whole piece of audio data are extracted;
in a preferred embodiment of the present invention, as shown in fig. 1, the abnormal sound detection module 3 further includes a preprocessing unit 34, which is respectively connected to the extracting unit 31 and the classifying unit 32, and is configured to perform normalized data processing on the extracted abnormal segment as an abnormal segment.
Because interference sounds, such as speaking sounds of staff, screwing sounds of maintenance equipment and the like, can exist in the environment when abnormal sound detection is carried out, the staff and the maintenance equipment need to be classified and removed; first, all the above extracted abnormal segments are normalized, and then mel-frequency cepstral coefficient (MFCC) feature extraction is performed, and MFCC (Mel Frequency Cepstral Coefficients) is a feature representation method commonly used for speech signal processing and speech recognition. The MFCC features are mainly used to extract spectral features in speech signals, which simulate the auditory properties of the human ear. Inputting the mel frequency cepstrum coefficients of each abnormal section into a classification model obtained by pre-training to obtain weight coefficients of each abnormal section, wherein different weight coefficients represent different sounds, so that the audio data of abnormal sounds of riding equipment are extracted, and the influence of environmental noise is removed;
if the number of abnormal segments exceeds the threshold value of the number of abnormal sounds and the occurrence frequency of the abnormal sounds is between 1 and 2hz (which is equivalent to the riding frequency of a user), the abnormal sounds which occur periodically are generated when the user rides, and the abnormal sounds are not detected.
In general, noise of riding equipment is detected through two aspects of noise detection and abnormal sound detection, so that accurate and automatic noise detection of the riding equipment with unified standards is realized.
The invention also provides a noise detection method of the riding equipment, which is applied to the noise detection system, as shown in fig. 2, and comprises the following steps:
step S1, a noise detection system collects sounds of detection points of riding equipment in real time in the riding detection process to obtain audio data;
s2, extracting audio characteristics of the audio data by the noise detection system, and carrying out noise detection according to the audio characteristics to obtain a noise detection result;
step S3, the noise detection system extracts abnormal segments from the audio data according to the audio characteristics, carries out sound classification on the abnormal segments, and carries out abnormal sound judgment according to the classification result to obtain an abnormal sound judgment result;
step S4, the noise detection system judges whether the noise detection result represents noise or whether the abnormal sound judgment result represents abnormal sound or not:
if yes, prompting the inspector that the riding equipment is not passed;
if not, prompting the detection personnel to pass the detection of the riding equipment.
In a preferred embodiment of the present invention, as shown in fig. 3, step S2 includes:
step S21, the noise detection system calculates A weight sound pressure level of each data frame in the audio data as corresponding audio characteristics;
step S22, the noise detection system determines whether the weighted sound pressure level a exceeds the preset maximum sound pressure level and the duration exceeds the longest duration:
if yes, the noise detection is not passed as a noise detection result;
if not, the noise detection is passed as a noise detection result.
In a preferred embodiment of the present invention, as shown in fig. 4, step S3 includes:
step S31, the noise detection system respectively extracts continuous data frames with the audio characteristics exceeding a preset abnormal value in the audio data as an abnormal section;
step S32, the noise detection system extracts the characteristics of each abnormal segment to obtain corresponding mel cepstrum coefficients, and inputs each mel cepstrum coefficient into a pre-trained classification model to obtain the sound classification result of each abnormal segment, and eliminates the abnormal segments except for riding equipment according to the sound classification result;
step S33, the noise detection system determines whether the number of remaining abnormal segments is greater than the abnormal sound number threshold and whether the time interval between adjacent abnormal segments is within a preset time range:
if yes, the abnormal sound detection is not passed as an abnormal sound detection result;
if not, the abnormal sound detection is used as an abnormal sound detection result.
In a preferred embodiment of the present invention, as shown in fig. 5, step S31 includes:
step S311, the noise detection system marks the data frame which exceeds the preset noise threshold and the previous data frame is not higher than the noise threshold as the recording starting point in the audio data;
step S312, the noise detection system marks the data frame with the first audio feature lower than the preset stopping value as the recording end point when the audio feature of the data frame is higher than the abnormal value after the recording start point;
in step S313, the noise detection system extracts each data frame between the recording start point and the recording end point, and a plurality of data frames before the recording start point and a plurality of data frames after the recording end point as one abnormal segment.
The foregoing is merely illustrative of the preferred embodiments of the present invention and is not intended to limit the embodiments and scope of the present invention, and it should be appreciated by those skilled in the art that equivalent substitutions and obvious variations may be made using the description and illustrations herein, which should be included in the scope of the present invention.
Claims (10)
1. A noise detection system of a riding device, wherein a flywheel of the riding device and/or a seat sleeve are set as detection points, comprising:
the sound acquisition device is used for acquiring the sound of the detection point of the riding equipment in real time in the riding detection process to obtain audio data;
the noise detection module is connected with the sound acquisition device and is used for extracting the audio characteristics of the audio data and carrying out noise detection according to the audio characteristics to obtain a noise detection result;
the abnormal sound detection module is connected with the noise detection module and is used for extracting abnormal sections from the audio data according to the audio characteristics, classifying the abnormal sections in sound and judging abnormal sound according to the classification result to obtain an abnormal sound judgment result;
and the qualification judging module is respectively connected with the noise detecting module and the abnormal sound detecting module and is used for prompting a detector that the riding equipment is not detected when the noise detecting result indicates noise or the abnormal sound judging result indicates abnormal sound.
2. The noise detection system of claim 1, wherein the noise detection module comprises:
the characteristic extraction unit is used for respectively calculating the A weight sound pressure level of each data frame in the audio data as the corresponding audio characteristic;
and the noise judgment unit is connected with the characteristic extraction unit and is used for not passing the noise detection as the noise detection result when the A weight sound pressure level exceeds a preset maximum sound pressure level and the duration exceeds the longest duration, and passing the noise detection as the noise detection result when the A weight sound pressure level does not exceed the preset maximum sound pressure level or the duration does not exceed the longest duration.
3. The noise detection system of claim 1, wherein the abnormal sound detection module comprises:
an extracting unit, configured to extract, as one of the anomaly segments, each of the continuous data frames in the audio data, where the audio feature of the data frame exceeds a preset anomaly value;
the classification unit is connected with the extraction unit and is used for extracting the characteristics of each abnormal section to obtain corresponding mel cepstrum coefficients, inputting each mel cepstrum coefficient into a pre-trained classification model to obtain a sound classification result of each abnormal section, and eliminating the abnormal sections except for the riding equipment according to the sound classification result;
and the abnormal sound judging unit is connected with the classifying unit and is used for not passing abnormal sound detection as an abnormal sound detection result when the number of the rest abnormal sections is judged to be larger than an abnormal sound number threshold value and the time interval between every two adjacent abnormal sections is in a preset time range, and passing abnormal sound detection as the abnormal sound detection result when the number of the abnormal sections is judged to be not larger than an abnormal sound number threshold value or the time interval between every two adjacent abnormal sections is judged to be not in the preset time range.
4. A noise detection system according to claim 3, wherein the extraction unit comprises:
a first extraction subunit, configured to mark, in the audio data, the data frame that exceeds a preset noise threshold and that is not higher than the noise threshold as a recording start point;
a second extraction subunit, connected to the first extraction subunit, configured to mark, as a recording end point, a data frame having a first audio feature lower than a preset stop value when the audio feature of the data frame after the recording start point is higher than the abnormal value;
and a third extraction subunit, connected to the first extraction subunit and the second extraction subunit, configured to extract, as one of the abnormal segments, each of the data frames between the recording start point and the recording end point, and a plurality of the data frames before the recording start point and a plurality of the data frames after the recording end point.
5. The noise detection system according to claim 3, wherein the abnormal sound detection module further comprises a preprocessing unit respectively connected to the extracting unit and the classifying unit, and configured to perform normalized data processing on the extracted abnormal segment as the abnormal segment.
6. The noise detection system according to claim 1, further comprising a man-machine interaction interface module respectively connected to the sound collection module, the noise detection module, the abnormal sound detection module, and the qualification judgment module, for displaying a waveform chart, a spectrogram, an energy chart, and the audio characteristics corresponding to the audio data, and the noise detection result and the abnormal sound judgment result.
7. A noise detection method for a riding device, applied to the noise detection system according to any one of claims 1 to 6, comprising:
step S1, the noise detection system collects the sound of a detection point of the riding equipment in real time in the riding detection process to obtain audio data;
step S2, the noise detection system extracts the audio characteristics of the audio data, and performs noise detection according to the audio characteristics to obtain a noise detection result;
step S3, the noise detection system extracts abnormal sections from the audio data according to the audio characteristics, carries out sound classification on the abnormal sections, and carries out abnormal sound judgment according to classification results to obtain abnormal sound judgment results;
step S4, the noise detection system judges whether the noise detection result represents noise or whether the abnormal sound judgment result represents abnormal sound or not:
if yes, prompting a detector that the riding equipment is not passed;
if not, prompting the detection personnel that the riding equipment passes the detection.
8. The method of detecting noise according to claim 7, wherein the step S2 includes:
step S21, the noise detection system calculates A weight sound pressure level of each data frame in the audio data as the corresponding audio feature;
step S22, the noise detection system determines whether the a weight sound pressure level exceeds a preset maximum sound pressure level and the duration exceeds a maximum duration:
if yes, the noise detection is not passed as the noise detection result;
if not, the noise detection is passed as the noise detection result.
9. The method of detecting noise according to claim 7, wherein the step S3 includes:
step S31, the noise detection system respectively extracts each continuous data frame with the audio characteristic exceeding a preset abnormal value in the audio data as one abnormal section;
step S32, the noise detection system performs feature extraction on each abnormal section to obtain a corresponding Mel cepstrum coefficient, respectively inputs each Mel cepstrum coefficient into a pre-trained classification model to obtain a sound classification result of each abnormal section, and eliminates the abnormal section except for the riding equipment according to the sound classification result;
step S33, the noise detection system determines whether the number of remaining abnormal segments is greater than an abnormal sound number threshold and whether a time interval between adjacent abnormal segments is within a preset time range:
if yes, the abnormal sound detection is not passed as the abnormal sound detection result;
if not, the abnormal sound detection is used as the abnormal sound detection result.
10. The method of detecting noise according to claim 9, wherein the step S31 includes:
step S311, the noise detection system marks, as a recording start point, the data frame that exceeds a preset noise threshold and that is not higher than the noise threshold in the audio data;
step S312, the noise detection system marks the data frame with the first audio feature lower than a preset stopping value as a recording end point when the audio feature of the data frame after the recording start point is higher than the abnormal value;
in step S313, the noise detection system extracts each of the data frames between the recording start point and the recording end point, and the plurality of data frames before the recording start point and the plurality of data frames after the recording end point as one of the abnormal segments.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311152024.4A CN117457026A (en) | 2023-09-07 | 2023-09-07 | Noise detection system and method for riding equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311152024.4A CN117457026A (en) | 2023-09-07 | 2023-09-07 | Noise detection system and method for riding equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117457026A true CN117457026A (en) | 2024-01-26 |
Family
ID=89588070
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311152024.4A Pending CN117457026A (en) | 2023-09-07 | 2023-09-07 | Noise detection system and method for riding equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117457026A (en) |
-
2023
- 2023-09-07 CN CN202311152024.4A patent/CN117457026A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101493552B1 (en) | Signal judgment method, signal judgment apparatus, program, and signal judgment system | |
| CN109034046B (en) | Method for automatically identifying foreign matters in electric energy meter based on acoustic detection | |
| US20110213612A1 (en) | Acoustic Signal Classification System | |
| CN106323452B (en) | A kind of detection method and detection device of equipment abnormal sound | |
| US20140149117A1 (en) | Method and system for identification of speech segments | |
| EP3166239B1 (en) | Method and system for scoring human sound voice quality | |
| CN107170465B (en) | Audio quality detection method and audio quality detection system | |
| CN108896878A (en) | A kind of detection method for local discharge based on ultrasound | |
| CN103546853A (en) | Speaker abnormal sound detecting method based on short-time Fourier transformation | |
| CN101976564A (en) | Method for identifying insect voice | |
| EP2881948A1 (en) | Spectral comb voice activity detection | |
| CN110568073B (en) | Method for picking up impact signal in noise environment | |
| CN112098939B (en) | Method and device for identifying and evaluating noise pollution source | |
| CN115410711B (en) | White feather broiler health monitoring method based on sound signal characteristics and random forest | |
| CN105513598A (en) | Playback voice detection method based on distribution of information quantity in frequency domain | |
| CN112331231A (en) | Broiler feed intake detection system based on audio technology | |
| CN112382313A (en) | Audio communication quality evaluation system and method | |
| Singh et al. | Linear Prediction Residual based Short-term Cepstral Features for Replay Attacks Detection. | |
| CN111276161B (en) | Voice quality scoring system and method | |
| CN113723521A (en) | Motor abnormal sound identification method and identification system | |
| JP4926588B2 (en) | Insulation discharge sound discrimination method and apparatus | |
| CN117457026A (en) | Noise detection system and method for riding equipment | |
| CN116364108A (en) | Transformer voiceprint detection method and device, electronic equipment and storage medium | |
| CN116230018A (en) | Synthetic voice quality evaluation method for voice synthesis system | |
| CN114566185A (en) | Abnormal sound automatic detection method and equipment and computer readable storage medium |
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 |