Irin Àwùjọ · 2023
Ensenas Ohun èlò Ìbánisọ̀rọ̀
by Saeli Seira Saenz and Alana Betanzo Carriles
- Àmì-ẹ̀yẹ Irin A' Design Award 2023 in Àwùjọ
- CLIENT: Universidad de Monterrey
- Estimated Number of Design Views: 34,628
- Verification code CCB43F0D2F · Verify this award
Colors
About Ensenas Ohun èlò Ìbánisọ̀rọ̀
Ensenas jẹ́ ohun èlò dígítà tí a ṣẹ̀dá láti mú ìbánisọ̀rọ̀ dára sí i láàrin àwọn tí kò lè gbọ́ràn àti àwọn tí ó lè gbọ́ràn ní Mẹ́síkò. Ensenas ń lo àwòṣe ẹ̀kọ́ ẹ̀rọ (machine learning) tí ó máa ń kà Èdè Àwọn Adití ti Mẹ́síkò (Mexican Sign Language) ní àkókò gidi, tí ó sì ń tú àwọn ọ̀rọ̀ sí èdè Sípáníìṣì tí a kọ sílẹ̀. A fi àkójọpọ̀ dátà ti fídíò 400 tí ó ní àwọn àmì tí a máa ń lò lóòrèkóòrè kọ́ àwòṣe náà, èyí tí a ti fi àmì sí tí a sì ti ṣàyẹ̀wò férémù sí férémù nípasẹ̀ ìṣírò ìdúró ara (pose estimation). Wẹ́ẹ̀bùsàìtì kan tí ó ṣe é lò dáadáa ní gbogbo ohun èlò (responsive website) ń ṣàfihàn iṣẹ́ náà, ó sì ń fúnni ní ìwádìí, àwòṣe àti àkójọpọ̀ dátà náà láti gbà sílẹ̀ lọ́fẹ̀ẹ́. Ète rẹ̀ ni láti dín àlàfo ìbánisọ̀rọ̀ náà kù.
About the Àmì-ẹ̀yẹ Irin A' Design Award
Ensenas Ohun èlò Ìbánisọ̀rọ̀ jẹ́ olùborí Àmì-ẹ̀yẹ Irin A' Design Award, ìdánimọ̀ olókìkí tí a fún àwọn àwòṣe tó dára tí wọ́n pàdé àwọn ohun tí iṣẹ́ àti ilé-iṣẹ́ nílò tí a retí láti ọ̀dọ̀ àwọn ẹ̀dá tí a ṣàwòṣe dáradára. Àwọn olùborí Irin ń fi ìjáfáfá ńlá, ìmọ̀ pàtó gíga, òye ògbóǹtarìgì àti agbára ìṣẹ̀dá àwọn olùṣẹ̀dá wọn hàn. A retí pé Ensenas Ohun èlò Ìbánisọ̀rọ̀ yóò jẹ́ ẹ̀dá tí a kà sí gidigidi, tí a gbóríyìn fún tí a sì nífẹ̀ẹ́, tó wúlò tó sì ní ìmúṣẹ̀dá tuntun. Àwòṣe tó borí Àmì-ẹ̀yẹ Irin A' Design Award yìí ń kó àwọn ọ̀nà tó dára jùlọ ilé-iṣẹ́ jọ, ó sì ní àwọn ànímọ́ ìmọ̀-ẹ̀rọ tó péye tí a so mọ́ ìmọ̀ ṣíṣe àti ẹ̀bùn. A bọ̀wọ̀ fún Ensenas Ohun èlò Ìbánisọ̀rọ̀ nítorí ìfọ̀kànbalẹ̀ rẹ̀, a sì ṣàwòṣe rẹ̀ láti pèsè ìmúdára ìgbésí ayé tó ń ran ayé lọ́wọ́ láti di ibi tó dára jù.
Press Kit
- Unique Properties
- Likewise, thanks to technology and current tools, we were able to generate a Machine Learning model or tool which makes a real time reading of the Mexican Sign Language; where it is possible to solve the communication gap between people with hearing disabilities and people without such disability; avoiding that people make the effort to learn the Mexican Sign Language.
- Dimensions & Packaging
- This is a model in a website, whit a responsive design.
- Tags
- IA, Machine Learning, LSM, Data base, Service Design, education and tools, Sign Language,
- Production Technology
- We developed a model and data set of machine learning in order to create new tools and opportunities for people with hearing disabilities and make their lives easier through it. We created a platform in which anyone can learn more about the project and download the machine learning model.
- Design Challenge
- The first step in creating this project was the creation of a data set consisting of a collection of 400 videos of the most used words in the Mexican sign language. These were ordered by categorize and label individually.
The next step was the pose prediction process, in which the model analyzes the content of the data set analyzing frame by frame by means of ML pipe pose Estimation.
At the end of this estimation, our supervised learning model is able to detect patterns in the poses without the need for human intervention in order to identify which words, letters, numbers and phrases they belong to. To complete this process, a manual selection of frames was also performed in order to achieve a more accurate prediction of the signals.
In order to open this project to the public, a digital platform was created in which it is possible to freely download our research, the LSM recognition model, the data set generated so that other developers, technology companies, educational institutions can take advantage of our project and can create more inclusion tools for the non-hearing community. - Research
- It is estimated that 5% of the world population, being 430 million people, will suffer a loss of disabling audio and in Mexico there are limited resources and tools to create an environment where everyone feels heard.
Following a design process, we conducted a series of interviews with experts, case studies and surveys to understand the perception of hearing impairment. We found that one of the limitations for the use of sign language is the learning curve.
We also found that some basic tasks for people with this disability can be more complicated than normal, which limits their independence. That's why we wanted to find a solution
inclusion to this problem.
This is Enseñas, an initiative that seeks inclusion through a digital tool made by through a machine learning model (LSM Model) with the purpose of creating a new way of
communication that facilitates the interaction between people with hearing disabilities and hearing people in real times. This model will be able to translate words from the Mexican sign language into Spanish written and vice versa in real time. - Inspiration
- The lack of education and tools about Mexican Sign Language to improve the communication and experience of hearing impaired people in Mexico, has led us to create awareness and promote the inclusion of this problem through a digital platform.
- Project Duration
- 6 months on Universidad de Montery and Monterrey City in México
