[Practical] Full record of mimie’s application to Gugen2017.

I applied to Gugen again this year, as I did last year.

I just received the notice of failure today, and I am still in shock, but I will keep this as a record before I forget.
And we will use it as food for the next year.

Promo video available on YouTube

About Gugen2017


This year we received 159 entries.

By the way, I want it! The first place in the number of votes was taken by a digital board game from the Nagoya Institute of Technology with 1,167 votes.

I'd love to have this piece as well! Looking at the 12 works with the highest number, 5 of them are ingenious with LEDs.
LEDs may seem simple, but they still seem to have a lot of depth.

competition entry

mimie \-"mimie"\-"mimie"\-"mimie"\-"mimie"\-"mimie"\-"mimie"\-"mimie"}- An elephant watches over children with sound.

A badge with a built-in microphone to watch over children with sound.
I took the letters MIMIE for ears.


  • Deep learning of voice characteristics through fun chats with children
  • Perform emotion recognition only when recognized as a child's voice (using empath API)
  • If you are feeling anxious, call your parent's phone.

reason for application

In my case, the biggest reason may be that I want to use high-quality software, although it may be a bit unique.

I come from a gamer background, and I love to improve my level of skill by gradually conquering difficult problems.
Furthermore, for some reason, I loved strategy books and would read them all day long once I picked out my favorite one.

Now that I am an adult, I don't play games anymore, but my personality hasn't changed from those days, and when I find a good software best manual, I get immersed in it and can't help but use it for something.

To be more specific, After Effects, Fusion360, and MATLAB are my favorite software at the highest level right now.

It is no exaggeration to say that I make things because I want to use these software programs.

In addition, the fact that Gugen is Japan's largest hardware contest and attracts a lot of attention, as well as the fact that it is just the right time to close out the year, matched my needs and led me to apply.

Motivation for the work

Last year it was a gadget for cocktail lovers, and this year we made it a gadget for parenting.

The reason why I chose to make a parenting story is because I wanted the inevitability of my making it.

As a lesson learned last year,

  • What I don't want doesn't resonate with me.

  • If there is no inevitability, you don't know what to make, and you are lost.

The following is a list of the most important

Therefore, this year we considered what to make with these in mind.

I came up with this item because my own child is now 3 years old and I was just thinking about how much more active he is becoming.

In addition, we spent quite a bit of time conceptualizing this year.

Of the roughly two and a half month period, I spent about a month and a half just thinking about it.

The elephant was chosen as the model because it is an animal that can hear so well that it can communicate with its friends 1 km away, and because its motherly nature fits the image of "watching over" the animals.


IMG 0003
IMG 6277
IMG 4754

I put my image into consideration for the hardware configuration, logo, system, etc.

big picture

I will write about how exactly I made them in separate articles, respectively.

Here is just a brief overview of the whole process.


We use CHAINER throughout.

I am fully indebted to docomo's API when I am in chat mode with my children.
We also use Empath and twilio's API when in watchful waiting mode.

And then there's the LINE bot, which is, in essence, a mishmash.

Mechanical Configuration

Hard composition

The blue body serves as the speaker and charging station, while the green face is the main one, with raspy, microphone, GPS, etc.

It was inevitably large because it was made by combining ready-made products.

If I had more time and energy, I would have liked to make something proper with KiCad.

Electrical Configuration

WatchOverBudge 03

I used Fritzing again because it is my favorite.

If you decide on pin assignments, modules, etc. as concretely as possible in Frizing before actually creating the configuration, subsequent work will go very smoothly.

speaker recognition

Deep learning with RNN on Raspi.

Using the code from the magazine Interface as a reference, I tweaked it by tweaking the layers a bit and increasing or decreasing the classifications.

We had a hard time getting the desired accuracy and it also took a long time to learn. We also had help from Sakura Server's VPS. We are still struggling here.

Filming and Video Production

The shooting took place in a neighborhood park. Nearby children looked at us quite suspiciously, w

The video was created with AfterEffects, my favorite software.

However, it was already less than three days before the application deadline at that time, so it was quite a rush job.

Next time I want to take a little more time to make it cooler,...


I managed to complete my application 10 minutes before the deadline!
Quite close to the edge,...

As a general rule of thumb, we learned that we need to have the product almost completed one month before the deadline in order to be satisfied with both the product and the video.

For next year

Next year, I want to move things using MATLAB and actuators that I haven't touched much this year!

We will be in input mode again for a while.