Started providing AI analysis module “SkySound” that simulates human auditory characteristics
Sky Disc Corporation Started providing AI analysis module “SkySound” that simulates human auditory characteristics Analysis of more than 10,000 feature data, AI can make judgments close to human ears in inspection work ………………………………………………………………………………………… SkyDisc Co., Ltd. (Head office: Fukuoka City, Fukuoka Prefecture; CEO: Takashi Hashimoto, hereinafter “the company”) is an AI analysis system specializing in the manufacturing industry as one of the analysis modules of “SkyAI (SkyAI)”. Succeeded in developing an analysis function “SkySound (Sky Sound)” that simulates auditory characteristics and started offering it. What is “SkySound”? “SkyAI” provided by our company is a service that provides an AI learning model created through shaping and analysis of data collected from IoT sensor devices etc. at manufacturing sites through API, and is incorporated into the system at the site. You can use it. We have developed “SkyModule” with the data shaping and analysis function of “SkyAI” as a module. As “SkyModule”, we are developing services that can be used in the manufacturing industry such as vibration analysis and factor analysis, but as the first installment, we have started offering “SkySound”, a sound analysis module.
[Image 1: https://prtimes.jp/i/2224/30/resize/d22401-30-116701-0.png]
Image of SkyAI and SkyModule In the analysis of sound data, the conventional method of FFT analysis (*) is often used, but it is a method that is largely different from the mechanism by which people understand the characteristics of sound. Therefore, the feature components of sounds that people are hearing to determine normality / abnormality and the features that appear as analysis results in the conventional FFT analysis are largely different, and the accuracy of the determination may be low depending on the type of sound. There was also. “SkySound” combines various feature extraction methods to create over 10,000 large amounts of feature data from one minute of sound data. The generated feature data is analyzed to the most effective features using our big data analysis technology. The combination of feature creation, refinement, and analysis techniques has made it possible to detect sound features that were difficult to capture with conventional methods. ※ FFT analysis: Abbreviation of Fast Fourier Transform analysis, also called fast Fourier transform, which transforms time series data of sound into frequency domain. It detects which frequency component is included in the time series data of sound.
[Image 2: https://prtimes.jp/i/2224/30/resize/d22401-3-698324-1.png]
Background of development of sound analysis module “SkySound” So far, we have provided an application service called “smart auscultation stick” that can judge the quality of machines by “sounds” recorded using the microphone function of a smartphone. This is a service that allows AI-equipped applications to determine the differences between the unique vibration sounds emitted from each device, which can only be discerned by skilled technicians who possess advanced technologies. I received it. However, there is a need for flexible use in every field, and by providing as a module via API instead of providing apps, it can be incorporated into the system in the field actually used by the customer, and high precision Developed a sound analysis module “SkySound” that can perform various analysis. ??How to Use 1) Check the site and consult on how to obtain the best sound data. 2) Acquire sound data for analysis. Acquire at least 100 sound data. 3) Create an evaluation AI using “SkySound”. 4) Perform accuracy verification using evaluation AI 5) Develop a system incorporating “SkySound”. 6) Operate using the developed system Usage example ・ Inspection work of finished products such as motor ・ Maintenance of equipment generating sound in factory ・ Striking inspection of concrete ※ Because it is also necessary to extract the feature value of abnormal data, it can be used only for the work that can acquire abnormal data. In addition, we incorporate this module into IoT products developed in-house and are also looking for partner companies that want to develop new AI products.
【Skydisk Inc. Company Profile】
Company Name: Sky Disc Co., Ltd. URL: skydisc.jp/ location: 2-3-6 Maizuru Chuo-ku, Fukuoka-shi, Fukuoka Prefecture Akasaka Prime Building 4F (Main Store) 4-2-11, Kudan Minami, Chiyoda-ku, Tokyo Avista Ichigaya Building 2F (Tokyo Head Office) Representative: Tsukasa Hashimoto (CEO) Established: October 1, 2013 Business Description: ・ Provision of AI × IoT service ・ Smart factory promotion promotion support ・ Providing in-house package
【Copyright / Trademark】
※ Company names and product names mentioned are registered trademarks and trademarks of their respective owners. ※ The information contained in this press release is current as of the announcement date. Please note that information such as product and service charges, service content and specifications, and contact information may be changed without notice.
For more information about this release(Japanese):
prtimes.jp/main/html/rd/p/000000030.000022401.html ※ This email is sent automatically, please refrain from reply. ━ ━ ━ ━