TWI825481B - Inspecting system and method for noise source, computer program product with stored programs, and computer readable medium with stored programs - Google Patents
Inspecting system and method for noise source, computer program product with stored programs, and computer readable medium with stored programs Download PDFInfo
- Publication number
- TWI825481B TWI825481B TW110134764A TW110134764A TWI825481B TW I825481 B TWI825481 B TW I825481B TW 110134764 A TW110134764 A TW 110134764A TW 110134764 A TW110134764 A TW 110134764A TW I825481 B TWI825481 B TW I825481B
- Authority
- TW
- Taiwan
- Prior art keywords
- model
- module
- data
- noise
- notebook computer
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims description 21
- 238000004590 computer program Methods 0.000 title claims description 9
- 238000013528 artificial neural network Methods 0.000 claims abstract description 42
- 238000012360 testing method Methods 0.000 claims abstract description 41
- 238000004458 analytical method Methods 0.000 claims abstract description 24
- 238000001228 spectrum Methods 0.000 claims abstract description 22
- 238000010183 spectrum analysis Methods 0.000 claims abstract description 18
- 238000001514 detection method Methods 0.000 claims description 75
- 238000012549 training Methods 0.000 claims description 40
- 238000010586 diagram Methods 0.000 claims description 18
- 238000012795 verification Methods 0.000 claims description 9
- 230000003287 optical effect Effects 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 230000005284 excitation Effects 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims description 2
- 239000000523 sample Substances 0.000 claims 9
- 238000007796 conventional method Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007717 exclusion Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Images
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
Description
本發明主要為一種檢測系統及方法，特別是有關於一種用以檢測造成筆記型電腦發出噪音的噪音來源的噪音源檢測系統、方法、內儲程式之電腦程式產品及內儲程式之電腦可讀取記錄媒體。 The present invention mainly relates to a detection system and method. In particular, it relates to a noise source detection system and method for detecting the noise source that causes the noise generated by a notebook computer, a computer program product of a stored program, and a computer readable program of the stored program. Take the recording medium.
按，筆記型電腦在設計與試產過程中，其內建喇叭在播放聲音當下，若與其他零組件產生共振效果時，該筆記型電腦就會有噪音問題的產生。習知檢測噪音源方法，係透過以人工交叉比對及反覆實驗驗證的方式，以找出造成該筆記型電腦產生噪音的噪音源來源，該習知檢測噪音源方法，係具有檢測時間過長及高人力成本的問題。 Press, during the design and trial production process of a notebook computer, if the built-in speaker resonates with other components when playing sound, the notebook computer will have noise problems. The conventional method of detecting noise sources is to find out the source of the noise source that causes the notebook computer to generate noise through manual cross-reference and repeated experimental verification. This conventional method of detecting noise sources has the disadvantage that the detection time is too long. and high labor costs.
有鑑於此，確實有必要提供一種噪音源檢測系統及方法，以解決上述之問題。 In view of this, it is indeed necessary to provide a noise source detection system and method to solve the above problems.
本發明的目的在於提供一種噪音源檢測系統及方法，係可以透過應用人工智慧技術，以自動檢測出造成筆記型電腦產生噪音的噪音源來源者。 The object of the present invention is to provide a noise source detection system and method that can automatically detect the source of the noise source that causes the notebook computer to generate noise by applying artificial intelligence technology.
本發明的次一目的在於提供一種內儲程式之電腦程式產品及內儲程式之電腦可讀取記錄媒體，用以執行上述方法。 A secondary object of the present invention is to provide a computer program product with a program stored therein and a computer-readable recording medium with a program stored therein to perform the above method.
為達成上述目的，本發明提供一種噪音源檢測系統，包含：一資料庫模組，用以儲存數個不同型號的筆記型電腦各自的一規格資訊，該規格資訊包含一機種型號、一喇叭數量及一喇叭配置位置，該資料庫模組另儲存數個第一影像及數個第二影像，以作為一訓練樣本資料，該數個第一影像為該數個筆記型電腦未產生噪音時的聲音頻譜圖，該數個第二影像為各該筆記型電腦在數種噪音源類型的影響下，相對應產生的聲音頻譜圖；一深度神經網路模組，電性連接該資料庫模組，將該訓練樣本資料作為輸入層資料，以及將該訓練樣本資料各自代表的筆記型電腦的機種型號、未產生噪音及產生噪音的噪音源類型作為輸出層資料，以訓練該深度神經網路模組；一檢測模組，電性連接該資料庫模組，該檢測模組用以檢測一待測筆記型電腦的喇叭數量，以及該待測筆記型電腦的喇叭配置位置，以獲得一檢測數據，該檢測模組將該檢測數據與儲存於該資料庫模組中各該筆記型電腦的規格資訊進行比對，以獲得一機種型號；一頻譜分析模組，用以感測該待測筆記型電腦的喇叭於一特定時間內所發出的聲波，以獲得一待測聲音頻譜圖；一顯示模組，用以顯示一輸出結果；及一處理器模組，電性連接該資料庫模組、該深度神經網路模組、該檢測模組、該頻譜分析模組及該顯示模組，該處理器模組控制該檢測模組及該頻譜分析模組作動，以取得該待測筆記型電腦的機種型號及待測聲音頻譜圖，該處理器模組將該待測聲音頻譜圖輸入至該深度神經網路模組，以取得一分析數據，該分析數據包含至少一筆記型電腦 的機種型號、該筆記型電腦未產生噪音的預測機率，以及各種噪音源類型造成該筆記型電腦產生噪音的預測機率，該處理器模組根據該待測筆記型電腦的機種型號，以由該分析數據中取得該輸出結果，該輸出結果為該待測筆記型電腦的機種型號，以及相對應該機種型號在於該數種噪音源類型的預測機率，以及未產生噪音的預測機率，該處理器模組將該輸出結果傳送並顯示於該顯示模組。 In order to achieve the above object, the present invention provides a noise source detection system, which includes: a database module used to store a specification information of several different models of notebook computers. The specification information includes a model model and a number of speakers. and a speaker configuration position. The database module also stores a plurality of first images and a plurality of second images as a training sample data. The plurality of first images are when the plurality of notebook computers do not generate noise. Sound spectrograms, the plurality of second images are corresponding sound spectrograms generated by each notebook computer under the influence of several types of noise sources; a deep neural network module is electrically connected to the database module , use the training sample data as the input layer data, and use the laptop computer models, non-noise-generating and noise-generating noise source types represented by the training sample data as the output layer data to train the deep neural network model. Set; a detection module, electrically connected to the database module, the detection module is used to detect the number of speakers of a notebook computer to be tested, and the speaker configuration position of the notebook computer to be tested, to obtain a detection data , the detection module compares the detection data with the specification information of each notebook computer stored in the database module to obtain a model model; a spectrum analysis module is used to sense the notebook to be tested The sound wave emitted by the speaker of the computer within a specific period of time is used to obtain a sound spectrum diagram to be measured; a display module is used to display an output result; and a processor module is electrically connected to the database module , the deep neural network module, the detection module, the spectrum analysis module and the display module, the processor module controls the operation of the detection module and the spectrum analysis module to obtain the notebook to be tested The model model of the computer and the sound spectrogram to be measured. The processor module inputs the sound spectrogram to be measured to the deep neural network module to obtain an analysis data. The analysis data includes at least one laptop computer. The model model, the predicted probability that the notebook computer does not produce noise, and the predicted probability that various noise source types cause the notebook computer to produce noise, the processor module is based on the model model of the notebook computer to be tested. The output result is obtained from the analysis data. The output result is the model model of the notebook computer to be tested, and the predicted probability that the model model is located in the several noise source types, and the predicted probability that no noise is generated. The processor model The group transmits and displays the output results on the display module.
本發明另提供一種噪音源檢測方法，包含：將數個不同型號的筆記型電腦，各自未產生噪音時的聲音頻譜圖，以及在數種噪音源類型的影響下相對應產生的聲音頻譜圖，作為一訓練樣本資料；該訓練樣本資料作為一深度神經網路模組的輸入層資料，以及，將該訓練樣本資料各自代表的筆記型電腦的機種型號、未產生噪音及產生噪音的噪音源類型作為該深度神經網路模組的輸出層資料，以訓練該深度神經網路模組；檢測一待測筆記型電腦的喇叭數量，以及該待測筆記型電腦的喇叭配置位置，以獲得一檢測數據；將該檢測數據與各該筆記型電腦的機種型號、喇叭數量及喇叭配置位置進行比對，以獲得一機種型號；感測該待測筆記型電腦的喇叭於一特定時間內所發出的聲波，以獲得一待測聲音頻譜圖；將該待測聲音頻譜圖輸入至該深度神經網路模組，以取得一分析數據，該分析數據包含至少一筆記型電腦的機種型號、該筆記型電腦未產生噪音的預測機率，以及各種噪音源類型造成該筆記型電腦產生噪音的預測機率；根據該待測筆記型電腦的機種型號，以由該分析數據中取得一輸出結果，該輸出結果為該待測筆記型電腦的機種型號，以及相對應該機種型號在於該數種噪音源類型的預測機率，以及未產生噪音的預測機率；及將該輸出結果顯示於一顯示模組。 The present invention also provides a noise source detection method, which includes: comparing the sound spectrograms of several different models of notebook computers when no noise is produced, and the corresponding sound spectrograms produced under the influence of several types of noise sources, As a training sample data; the training sample data is used as the input layer data of a deep neural network module, and the model and model of the notebook computer represented by the training sample data, and the type of noise source that does not generate noise and generates noise. As the output layer data of the deep neural network module, to train the deep neural network module; detect the number of speakers of a notebook computer to be tested, and the speaker configuration position of the notebook computer to be tested, to obtain a detection data; compare the detection data with the model number, number of speakers and speaker configuration positions of each laptop computer to obtain a model number; sense the sound emitted by the speakers of the laptop computer under test within a specific period of time sound waves to obtain a spectrogram of the sound to be measured; input the spectrogram of the sound to be measured into the deep neural network module to obtain an analysis data, the analysis data includes at least one model of the notebook computer, the notebook model The predicted probability that the computer does not produce noise, and the predicted probability that various noise source types cause the notebook computer to produce noise; according to the model model of the notebook computer to be tested, an output result is obtained from the analysis data, and the output result is The model model of the notebook computer to be tested, and the predicted probability of the several noise source types corresponding to the model model, and the predicted probability of no noise being generated; and displaying the output result on a display module.
在一些實施例中，該深度神經網路模組係能夠以深度學習的深度神經網路或卷積神經網路進行訓練，以及分析該待測聲音頻譜圖及輸出該分析數據。 In some embodiments, the deep neural network module is capable of training with a deep neural network or a convolutional neural network of deep learning, analyzing the spectrogram of the sound to be measured, and outputting the analysis data.
在一些實施例中，該檢測模組係可以為一自動光學檢測儀。 In some embodiments, the detection module may be an automatic optical detector.
在一些實施例中，該深度神經網路模組用以建構一測試模型，該測試模型的預配置參數可以分別為學習率等於0.001，權重衰減等於0.001，迭代次數等於2000，批量大小等於64，以及採用Leaky ReLU作為激勵函數。 In some embodiments, the deep neural network module is used to construct a test model. The preconfigured parameters of the test model may be the learning rate equal to 0.001, the weight decay equal to 0.001, the number of iterations equal to 2000, and the batch size equal to 64. And use Leaky ReLU as the excitation function.
在一些實施例中，該訓練樣本資料中的該數個不同型號的筆記型電腦，各自未產生噪音時的聲音頻譜圖與該數種噪音源類型的影響下相對應產生的聲音頻譜圖，隨機分配為一訓練資料、一測試資料及一驗證資料，該訓練資料、該測試資料及該驗證資料的筆數比例為7：2：1。 In some embodiments, the sound spectrograms of the notebook computers of different models in the training sample data when no noise is generated and the corresponding sound spectrograms produced under the influence of the several noise source types are randomly determined. It is allocated as one training data, one testing data and one verification data, and the ratio of the number of training data, testing data and verification data is 7:2:1.
本發明揭示之內儲程式之電腦程式產品及內儲程式之電腦可讀取記錄媒體，當電腦系統載入該程式並執行後，可完成上述方法；如此，可便於使用、交換或執行上揭方法，有利於廣泛運用上述的噪音源檢測方法於其他應用軟體。 The invention discloses a computer program product containing a program and a computer-readable recording medium containing the program. When the computer system loads the program and executes it, the above method can be completed; in this way, the above method can be easily used, exchanged or executed. This method is conducive to the widespread use of the above noise source detection method in other application software.
本發明噪音源檢測系統及方法具有下列特點：係可以透過檢測模組根據該待測筆記型電腦的喇叭數量及其喇叭配置位置，以得知該待測筆記型電腦的機種型號，以及，透過該頻譜分析模組獲得該待測筆記型電腦的待測聲音頻譜圖，該處理器模組將該待測聲音頻譜圖輸入至該深度神經網路模組，並根據該待測筆記型電腦的機種型號，並可得出該待測筆記型電腦在於該數種噪音源類型的預測機率，以及為產生噪音的預測機率。如此，當該筆記型電腦發出噪音時，設計及工程人員即可優先檢查預測機率最高的噪音源類型，以快速且精準的 找出噪音源，本發明噪音源檢測系統及方法，係可以達到提升偵錯時效性及減少人力及物力資源成本之功效。 The noise source detection system and method of the present invention have the following characteristics: the detection module can be used to learn the model number of the notebook computer to be tested based on the number of speakers and the configuration position of the speakers of the notebook computer to be tested, and, through The spectrum analysis module obtains the sound spectrogram of the notebook computer to be tested, and the processor module inputs the sound spectrogram to be measured into the deep neural network module, and analyzes the sound spectrum of the notebook computer under test based on the frequency spectrum of the notebook computer to be tested. Model model, and the predicted probability of the notebook computer under test being exposed to these types of noise sources and the predicted probability of generating noise can be obtained. In this way, when the laptop makes noise, designers and engineers can prioritize the type of noise source with the highest predicted probability to quickly and accurately detect To identify the noise source, the noise source detection system and method of the present invention can achieve the effect of improving the timeliness of debugging and reducing the cost of human and material resources.
1:資料庫模組 1: Database module
2:深度神經網路模組 2: Deep neural network module
21:測試模型 21: Test model
3:檢測模組 3: Detection module
4:頻譜分析模組 4: Spectrum analysis module
5:顯示模組 5:Display module
6:處理器模組 6: Processor module
S1:模型訓練步驟 S1: Model training steps
S2:機種檢測步驟 S2: Model detection steps
S3:聲音感測步驟 S3: Sound sensing step
S4:噪音源分析步驟 S4: Noise source analysis steps
S5:輸出結果步驟 S5: Output result step
[圖1]為本發明噪音源檢測系統的系統方塊圖；[圖2a]為本發明噪音源檢測系統之噪音源類型為未產生噪音的聲音頻譜圖；[圖2b]為本發明噪音源檢測系統之噪音源類型為蓋板未鎖緊螺絲的聲音頻譜圖；[圖2c]為本發明噪音源檢測系統之噪音源類型為主機板未鎖緊螺絲的聲音頻譜圖；[圖2d]為本發明噪音源檢測系統之噪音源類型為風扇未鎖緊螺絲的聲音頻譜圖；[圖2e]為本發明噪音源檢測系統之噪音源類型為金屬支架未鎖緊螺絲的聲音頻譜圖；[圖2f]為本發明噪音源檢測系統之噪音源類型為喇叭未裝設緩衝墊的聲音頻譜圖；[圖2g]為本發明噪音源檢測系統之噪音源類型為內部配線浮動的聲音頻譜圖；[圖3]為本發明噪音源檢測方法的步驟流程圖。 [Figure 1] is a system block diagram of the noise source detection system of the present invention; [Figure 2a] is a sound spectrum diagram of the noise source detection system of the present invention when the noise source type is no noise; [Figure 2b] is the noise source detection system of the present invention The sound spectrum diagram of the system when the noise source type is the unlocked screws of the cover plate; [Figure 2c] is the sound spectrum diagram of the noise source detection system of the present invention when the noise source type is the motherboard unlocked screws; [Figure 2d] is the sound spectrum diagram of the system. [Figure 2e] is a sound spectrum diagram of the noise source detection system of the invention where the noise source type is the fan without tightening the screws; [Figure 2e] is the sound spectrum diagram of the noise source detection system of the invention where the noise source type is the metal bracket without tightening screws; [Figure 2f ] is a sound spectrum diagram of the noise source detection system of the present invention when the noise source type is a speaker without a buffer pad; [Fig. 2g] is a sound spectrum diagram of the noise source detection system of the present invention when the noise source type is internal wiring floating; [Fig. 3] is a step flow chart of the noise source detection method of the present invention.
茲配合圖式將本發明實施例詳細說明如下，其所附圖式主要為簡化之示意圖，僅以示意方式說明本發明之基本結構，因此在該等圖式中僅標示與 本發明有關之元件，且所顯示之元件並非以實施時之數目、形狀、尺寸比例等加以繪製，其實際實施時之規格尺寸實為一種選擇性之設計，且其元件佈局形態有可能更為複雜。 The embodiments of the present invention are described in detail below with reference to the drawings. The accompanying drawings are mainly simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner. Therefore, only the symbols and The components related to the present invention, and the components shown are not drawn with the number, shape, size ratio, etc. of the implementation. The specifications and dimensions of the actual implementation are actually a selective design, and the layout of the components may be more complex.
以下各實施例的說明是參考附加的圖式，用以例示本發明可據以實施的特定實施例。本發明所提到的方向用語，例如「上」、「下」、「前」、「後」、等，僅是參考附加圖式的方向。因此，使用的方向用語是用以說明及理解本申請，而非用以限制本申請。另外，在說明書中，除非明確地描述為相反的，否則詞語“包含”將被理解為意指包含所述元件，但是不排除任何其它元件。 The following description of the embodiments refers to the accompanying drawings to illustrate specific embodiments in which the invention may be practiced. The directional terms mentioned in the present invention, such as "up", "down", "front", "back", etc., are only for reference to the directions in the attached drawings. Therefore, the directional terms used are used to explain and understand the present application, rather than to limit the present application. Additionally, in the specification, unless expressly described to the contrary, the word "comprising" will be understood to mean the inclusion of stated elements but not the exclusion of any other elements.
本發明全文所述之「電腦(Computer)」，係指具備特定功能且以硬體或硬體與軟體實現的各式資料處理裝置，例如：伺服器、虛擬機器(如：Amazon,Azure)、桌上型電腦、筆記型電腦、平板電腦或智慧型手機等，係本發明所屬技術領域中具有通常知識者可以理解。 "Computer" as mentioned throughout the present invention refers to various data processing devices that have specific functions and are implemented by hardware or hardware and software, such as servers, virtual machines (such as Amazon, Azure), Desktop computers, notebook computers, tablet computers, smart phones, etc. can be understood by those with ordinary knowledge in the technical field to which the present invention belongs.
本發明全文所述之「電腦程式產品(Computer Program Product)」，係指載有電腦可讀取之程式且不限外在形式之物，係本發明所屬技術領域中具有通常知識者可以理解。 The "Computer Program Product" mentioned in the entire text of this invention refers to something that contains a computer-readable program and is not limited to external form. It can be understood by those with ordinary knowledge in the technical field to which this invention belongs.
本發明全文所述之「電腦可讀取記錄媒體(Computer Readable Medium)」，係指一載體，其上儲存有軟體，該軟體可為電腦所存取；常見者有光碟、硬碟、USB隨身碟、各式記憶卡、雲端或虛擬儲存空間等，係本發明所屬技術領域中具有通常知識者可以理解。 The "Computer Readable Medium" mentioned in the full text of this invention refers to a carrier on which software is stored, and the software can be accessed by a computer; common examples include optical discs, hard drives, and USB pen drives. Disks, various memory cards, cloud or virtual storage spaces, etc., can be understood by those with ordinary knowledge in the technical field to which the present invention belongs.
請參照圖1所示，其係本發明噪音源檢測系統的較佳實施例，包含：一資料庫模組1、一深度神經網路模組2、一檢測模組3、一頻譜分析模組4、一顯示模組5及一處理器模組6，該資料庫模組1電性連接該深度神經網路
模組2及該檢測模組3，該資料庫模組1、該深度神經網路模組2、該檢測模組3及該頻譜分析模組4分別電性連接該處理器模組6。
Please refer to Figure 1, which is a preferred embodiment of the noise source detection system of the present invention, including: a database module 1, a deep
該資料庫模組1用以儲存數個不同型號的筆記型電腦各自的一規格資訊，該規格資訊包含一機種型號、一喇叭數量及一喇叭配置位置。在本實施例中，該機種型號可以包含A型號及B型號，其中，A型號的喇叭數量為2個，且設置於筆記型電腦的鍵盤兩側；B型號的喇叭數量為4個，且設置於筆記型電腦的鍵盤四個角落，惟不以此作為本發明的限制。 The database module 1 is used to store a specification information of several notebook computers of different models. The specification information includes a model model, a speaker quantity and a speaker configuration position. In this embodiment, the model may include model A and model B. Among them, the number of speakers of model A is 2 and they are arranged on both sides of the keyboard of the notebook computer; the number of speakers of model B is 4 and they are arranged At the four corners of the keyboard of the notebook computer, this is not a limitation of the present invention.
請一併參照圖2a~2g所示，該資料庫模組1另儲存數個第一影像及數個第二影像，以作為一訓練樣本資料。該數個第一影像為該數個筆記型電腦未產生噪音時的聲音頻譜圖，該數個第二影像為各該筆記型電腦在數種噪音源類型的影響下，相對應產生的聲音頻譜圖。在本實施例中，造成該筆記型電腦產生噪音的噪音源類型，係可以包含未鎖緊螺絲、喇叭未裝設緩衝墊及內部配線浮動等噪音源類型，例如但不限制地，該未鎖緊螺絲可以包含用以鎖固於筆記型電腦之蓋板、主機板、風扇及金屬支架的螺絲。 Please refer to Figures 2a to 2g. The database module 1 also stores several first images and several second images as a training sample data. The first images are the sound spectrum diagrams of the notebook computers when no noise is generated, and the second images are the sound spectrum diagrams produced by the notebook computers under the influence of various types of noise sources. Figure. In this embodiment, the types of noise sources that cause the notebook computer to generate noise may include unlocked screws, speakers without buffer pads, and floating internal wiring. For example, but not limited to, the unlocked Tightening screws may include screws used to secure the cover, motherboard, fan and metal bracket of the notebook computer.
該深度神經網路模組2電性連接該資料庫模組1，將該訓練樣本資料作為輸入層資料，以及將該訓練樣本資料各自代表的筆記型電腦的機種型號、未產生噪音及產生噪音的噪音源類型作為輸出層資料，以訓練該深度神經網路模組2。值得一提的是，圖2b~2g中所框選區域或是箭頭所指引位置，係為代表各種不同噪音源類型的特徵所在。
The deep
在本實施例中，該深度神經網路模組2可以採用PyTorch框架，以建構一測試模型21，該測試模型21的預配置參數可以如下列表一所示。該深度神經網路模組2將該訓練樣本資料作為該測試模型21的輸入層資料，以
及，將該訓練樣本資料各自代表的筆記型電腦的機種型號，以及未產生噪音及產生噪音的噪音源類型等預測結果，作為該測試模型21的輸出層資料，以訓練該測試模型21。
In this embodiment, the deep
詳言之，該測試模型21可以採用深度神經網路(DNN)或卷積神經網路(CNN)作為基礎架構，並使用監督式學習的訓練方式，該測試模型21各層的神經元權重及運算過程，係屬本發明相關領域中的通常知識，在此不多加贅述。在本實施例中，該測試模型21可以由官方網站下載使用，且其相關設定及權重等檔案亦可以直接採用官方網站所提供提資料。
In detail, the
該測試模型21在初次訓練時，各該筆記型電腦未產生噪音時的聲音頻譜圖的影像數量，至少需要167張，各該筆記型電腦各自在於不同噪音源類型所對應產生的聲音頻譜圖的影像數量，至少需要291張。藉此，該測試模型21可以自動學習如何辨識出筆記型電腦是否有產生噪音，且進一步辨識出產生噪音的噪音源類型。上數影像張數僅為範例，而非作為本發明的限制。
During the initial training of the
較佳地，該訓練樣本資料可以切割成訓練資料及測試資料，更進一步地，該訓練資料還可以切割出驗證資料，以對該測試模型21進行訓練及驗
證，以調整該測試模型21的神經網路權重。在本實施例中，該訓練資料、該測試資料及該驗證資料的比例可以為7：2：1。
Preferably, the training sample data can be cut into training data and test data. Furthermore, the training data can also be cut into verification data to train and verify the
該檢測模組3電性連接該資料庫模組1，該檢測模組3用以檢測一待測筆記型電腦的喇叭數量，以及該待測筆記型電腦的喇叭配置位置，以獲得一檢測數據。該檢測模組3將該檢測數據與儲存於該資料庫模組1中各該筆記型電腦的規格資訊進行比對，以獲得一機種型號。在本實施例中，該檢測模組3係為一自動光學檢測儀(AOI)，該自動光學檢測裝置用以檢測取得筆記型電腦的檢測數據及機種型號之相關技術，係屬本發明相關領域中具有通常知識者可以理解，在此不多加贅述。
The
該頻譜分析模組4用以感測該待測筆記型電腦的喇叭於一特定時間所發出的聲波，以獲得一待測聲音頻譜圖。該頻譜分析模組4可以為習知音頻頻譜分析儀，該習知音頻頻譜分析儀係可以為透過麥克風攫取聲音，並產生相對應的聲音頻譜圖的裝置。
The
該顯示模組5用以顯示一輸出結果，在本實施例中，該顯示模組5為一電腦螢幕或一行動裝置之顯示器。
The
該處理器模組6電性連接該資料庫模組1、該深度神經網路模組2、該檢測模組3及該頻譜分析模組4，該處理器模組6控制該檢測模組3及該頻譜分析模組4作動，以取得該待測筆記型電腦的機種型號及待測聲音頻譜圖。該處理器模組6將該待測聲音頻譜圖輸入至該深度神經網路模組2，以取得一分析數據，該分析數據包含至少一筆記型電腦的機種型號、該筆記型電腦未產生噪音的預測機率，以及各種噪音源類型造成該筆記型電腦產生噪音的預測機率。
The
該處理器模組6根據該待測筆記型電腦的機種型號，以由該分析數據中取得該輸出結果，該輸出結果為該待測筆記型電腦的機種型號，以及相對應該機種型號在於該數種噪音源類型的預測機率，以及未產生噪音的預測機率。該處理器模組6將該輸出結果傳送並顯示於該顯示模組5。藉此，當該筆記型電腦發出噪音時，設計及工程人員即可優先檢查預測機率最高的噪音源類型，以提升找到噪音源的機率，進而減少整體檢測時間。舉例而言，該筆記型電腦的機種型號，以及未產生噪音的預測機率及產生噪音的噪音源類型的預測機率可以如下列表二所示：
請參照圖3所示，其係本發明噪音源檢測方法的較佳實施例，包含：一模型訓練步驟S1，一機種檢測步驟S2，一聲音感測步驟S3、一噪音源分析步驟S4及一輸出結果步驟S5。 Please refer to Figure 3, which is a preferred embodiment of the noise source detection method of the present invention, including: a model training step S1, a model detection step S2, a sound sensing step S3, a noise source analysis step S4 and a Output the result step S5.
該模型訓練步驟S1用以將數個不同型號的筆記型電腦，各自未產生噪音時的聲音頻譜圖，以及在數種噪音源類型的影響下相對應產生的聲音頻譜圖，作為一訓練樣本資料。隨後，將該訓練樣本資料作為一深度神經網路模組的輸入層資料，以及，將該訓練樣本資料各自代表的筆記型電腦的機種型號、未產生噪音及產生噪音的噪音源類型作為該深度神經網路模組的輸出層資料，以訓練該深度神經網路模組。 The model training step S1 is used to use the sound spectrograms of several different models of notebook computers when no noise is produced, and the corresponding sound spectrograms produced under the influence of several types of noise sources, as a training sample data . Subsequently, the training sample data is used as the input layer data of a deep neural network module, and the model model of the laptop computer represented by the training sample data, and the type of noise source that does not generate noise and generates noise is used as the depth The output layer data of the neural network module to train the deep neural network module.
該機種檢測步驟S2用以檢測一待測筆記型電腦的喇叭數量，以及該待測筆記型電腦的喇叭配置位置，以獲得一檢測數據。隨後，將該檢測數據與各該筆記型電腦的機種型號、喇叭數量及喇叭配置位置進行比對，以獲得一機種型號。其中，該檢測數據係可以透過上述檢測模組3取得，惟不以此為限。
The model detection step S2 is used to detect the number of speakers of a notebook computer to be tested and the speaker configuration position of the notebook computer to be tested to obtain a detection data. Subsequently, the detection data is compared with the model number, number of speakers, and speaker configuration position of each notebook computer to obtain a model number. The detection data can be obtained through the above-mentioned
該聲音感測步驟S3用以感測該待測筆記型電腦的喇叭於一特定時間內所發出的聲波，以獲得一待測聲音頻譜圖。其中，該待測聲音頻譜圖係可以透過上述頻譜分析模組4取得，惟不以此為限。
The sound sensing step S3 is used to sense the sound waves emitted by the speaker of the notebook computer to be tested within a specific period of time to obtain a spectrogram of the sound to be tested. The spectrogram of the sound to be measured can be obtained through the above-mentioned
該噪音源分析步驟S4用以將該待測聲音頻譜圖輸入至該深度神經網路模組，以取得一分析數據。該分析數據包含至少一筆記型電腦的機種型號、該筆記型電腦未產生噪音的預測機率，以及各種噪音源類型造成該筆記型電腦產生噪音的預測機率。隨後，該噪音源分析步驟S4根據該待測筆記型電腦的機種型號，以由該分析數據中取得一輸出結果，該輸出結果為該待測筆記型電腦的機種型號，以及相對應該機種型號在於該數種噪音源類型的預測機率，以及未產生噪音的預測機率。 The noise source analysis step S4 is used to input the spectrogram of the sound to be measured into the deep neural network module to obtain an analysis data. The analysis data includes at least one laptop model, the predicted probability that the laptop does not generate noise, and the predicted probability that various noise source types cause the laptop to generate noise. Subsequently, the noise source analysis step S4 obtains an output result from the analysis data according to the model model of the notebook computer to be tested. The output result is the model model of the notebook computer to be tested, and the corresponding model model is The predicted probabilities of these noise source types, and the predicted probabilities of no noise being generated.
該輸出結果步驟S5用以將該輸出結果顯示於上述顯示模組5。
The output result step S5 is used to display the output result on the above-mentioned
本發明上述方法實施例還可利用程式語言(Program Language，如：C++、Java、Python或Julia)撰寫成電腦程式(如：噪音源檢測程式，用以辨識筆記型電腦的噪音源類型)，其程式碼的撰寫方式係熟知該項技藝者可以理解，可用以產生一種內儲程式之電腦程式產品，當該電腦系統載入該程式並執行後，可完成本發明上述方法實施例。 The above method embodiments of the present invention can also be written into a computer program (such as a noise source detection program to identify the type of noise source of a notebook computer) using a program language (Program Language, such as C++, Java, Python or Julia). The programming method of the program code can be understood by those familiar with the art, and can be used to generate a computer program product with a built-in program. When the computer system loads the program and executes it, the above-mentioned method embodiments of the present invention can be completed.
上述電腦程式產品還可儲存於一種內儲程式之電腦可讀取記錄媒體，如：光碟、硬碟、USB隨身碟、各式記憶卡、雲端或虛擬儲存空間等，當電腦系統載入上述程式並執行後，可完成本發明上述方法實施例，作為本發明之電腦系統軟硬體協同運作的依據。 The above-mentioned computer program products can also be stored in a computer-readable recording medium with built-in programs, such as optical discs, hard drives, USB flash drives, various memory cards, clouds or virtual storage spaces, etc. When the computer system loads the above-mentioned programs After execution, the above-mentioned method embodiments of the present invention can be completed, which serve as the basis for the cooperative operation of the computer system software and hardware of the present invention.
承上所述，本發明噪音源檢測系統及方法，係可以透過檢測模組根據該待測筆記型電腦的喇叭數量及其喇叭配置位置，以得知該待測筆記型電腦的機種型號，以及，透過該頻譜分析模組獲得該待測筆記型電腦的待測聲音頻譜圖，該處理器模組將該待測聲音頻譜圖輸入至該深度神經網路模組，並根據該待測筆記型電腦的機種型號，並可得出該待測筆記型電腦在於該數種噪音源類型的預測機率，以及為產生噪音的預測機率。如此，當該筆記型電腦發出噪音時，設計及工程人員即可優先檢查預測機率最高的噪音源類型，以快速且精準的找出噪音源，本發明噪音源檢測系統及方法，係可以達到提升偵錯時效性及減少人力及物力資源成本之功效。 Following the above, the noise source detection system and method of the present invention can use the detection module to learn the model and model of the notebook computer to be tested based on the number of speakers and the configuration position of the speakers of the notebook computer to be tested, and , through the spectrum analysis module, the sound spectrum diagram of the notebook computer to be tested is obtained, and the processor module inputs the sound spectrum diagram to be measured into the deep neural network module, and based on the notebook computer to be tested The model model of the computer can be used to obtain the predicted probability of the notebook computer under test being exposed to these types of noise sources, as well as the predicted probability of generating noise. In this way, when the notebook computer makes noise, the designers and engineers can check the noise source type with the highest predicted probability first to quickly and accurately find the noise source. The noise source detection system and method of the present invention can achieve improvement. The effectiveness of debugging timeliness and reducing human and material resource costs.
上述揭示的實施形態僅例示性說明本發明之原理、特點及其功效，並非用以限制本發明之可實施範疇，任何熟習此項技藝之人士均可在不違背本發明之精神及範疇下，對上述實施形態進行修飾與改變。任何運用本發明所揭示內容而完成之等效改變及修飾，均仍應為下述之申請專利範圍所涵蓋。 The above-disclosed embodiments are only illustrative of the principles, characteristics and effects of the present invention, and are not intended to limit the scope of the present invention. Anyone skilled in the art can implement the invention without violating the spirit and scope of the present invention. Modifications and changes are made to the above embodiments. Any equivalent changes and modifications accomplished by applying the contents disclosed in the present invention shall still be covered by the following patent application scope.
1:資料庫模組 1: Database module
2:深度神經網路模組 2: Deep neural network module
21:測試模型 21: Test model
3:檢測模組 3: Detection module
4:頻譜分析模組 4: Spectrum analysis module
5:顯示模組 5:Display module
6:處理器模組 6: Processor module
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW110134764A TWI825481B (en) | 2021-09-17 | 2021-09-17 | Inspecting system and method for noise source, computer program product with stored programs, and computer readable medium with stored programs |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW110134764A TWI825481B (en) | 2021-09-17 | 2021-09-17 | Inspecting system and method for noise source, computer program product with stored programs, and computer readable medium with stored programs |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| TW202314685A TW202314685A (en) | 2023-04-01 |
| TWI825481B true TWI825481B (en) | 2023-12-11 |
Family
ID=86943408
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| TW110134764A TWI825481B (en) | 2021-09-17 | 2021-09-17 | Inspecting system and method for noise source, computer program product with stored programs, and computer readable medium with stored programs |
Country Status (1)
| Country | Link |
|---|---|
| TW (1) | TWI825481B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0537967A (en) * | 1991-07-25 | 1993-02-12 | Sharp Corp | Noise inspecting device |
| CN111031490A (en) * | 2019-11-19 | 2020-04-17 | 长虹美菱股份有限公司 | Method for detecting vibration noise source of residential building |
| TW202025142A (en) * | 2018-12-21 | 2020-07-01 | 日商三菱電機股份有限公司 | Sound source direction estimation device, sound source direction estimation method and sound source direction estimation program |
-
2021
- 2021-09-17 TW TW110134764A patent/TWI825481B/en active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0537967A (en) * | 1991-07-25 | 1993-02-12 | Sharp Corp | Noise inspecting device |
| TW202025142A (en) * | 2018-12-21 | 2020-07-01 | 日商三菱電機股份有限公司 | Sound source direction estimation device, sound source direction estimation method and sound source direction estimation program |
| CN111031490A (en) * | 2019-11-19 | 2020-04-17 | 长虹美菱股份有限公司 | Method for detecting vibration noise source of residential building |
Also Published As
| Publication number | Publication date |
|---|---|
| TW202314685A (en) | 2023-04-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9697108B2 (en) | System, method, and apparatus for automatic recording and replaying of application executions | |
| US9454454B2 (en) | Memory leak analysis by usage trends correlation | |
| US9703670B2 (en) | Performance state machine control with aggregation insertion | |
| Chowdhury et al. | Greenscaler: training software energy models with automatic test generation | |
| JP2019075172A (en) | Identifying content in electronic images | |
| US9436449B1 (en) | Scenario-based code trimming and code reduction | |
| CN102306117B (en) | Automatic hardware test alarm method and device | |
| US9836876B1 (en) | Rendering images using ray tracing with multiple light sources | |
| CN111950279B (en) | Entity relationship processing method, device, equipment and computer readable storage medium | |
| US20170168913A9 (en) | Method and system for testing of applications in asset management software | |
| WO2021004329A1 (en) | Method and device for generating user information view | |
| US20150143342A1 (en) | Functional validation of software | |
| US20160358040A1 (en) | System and methods for trajectory pattern recognition | |
| CN114579452A (en) | Interface influence detection method and related equipment | |
| TWI825481B (en) | Inspecting system and method for noise source, computer program product with stored programs, and computer readable medium with stored programs | |
| Haq et al. | The popular tools of data sciences: Benefits, challenges and applications | |
| CN116860608A (en) | Interface testing method and device, computing equipment and storage medium | |
| US20170140080A1 (en) | Performing And Communicating Sheet Metal Simulations Employing A Combination Of Factors | |
| Chang et al. | MLHarness: A scalable benchmarking system for MLCommons | |
| CN113051153B (en) | Application software vulnerability scanning method and related equipment | |
| US10803219B1 (en) | Method and system for combined formal static analysis of a design code | |
| US20200013061A1 (en) | System and method for measuring user behavior consistency based on complex correspondence system | |
| CN117319091B (en) | Enterprise software network security vulnerability detection method and system based on deep learning | |
| CN114513736B (en) | Acoustic testing method, equipment, terminal and storage medium for earphone | |
| Huang et al. | Analyzing GUI running fluency for android apps |