Background Image


mirror mirror on the wall

when 2017

the problem

Day 1 - Look what the cat dragged in from the street

Found an old monitor from Acer (Model AL1716) in the streets for free. A backup monitor is always welcome so I took him home with me.

Day 2 - Again?

Girlfriends asks me in the morning when she is getting ready to work while I can stay in bed for another hour what time it is and when the next tram will leave from the nearest station...

Day 3 - Is this groundhog day?

Girlfriends asks me again in the morning what time it is and when the next tram will leave from the nearest station. Twice this time because she is in such a hurry...

Day 99 - Enough is Enough!

Girlfriends asks me like everyday in the morning several time what time it is and when the next tram will leave from the nearest station...

the solution

Day 100 - Help is on the way

I open my regular computer magazine the c't and find an article about on how you can build yourself a magic mirror that is both mirror and display. Powered by a RaspberryPI you can show all kinds of information on it like the weather, a clock and even other modules since it is open-source. Is there a module for displaying the local transportation system? Yes, there is and there is still this old monitor somewhere around here so let's get started.

Neat, I still have a RaspberryPI Model 1 too so lets format the sd card, install the standard OS and follow the instruction on the projects homepage.

Damn, the installation process won't finish, since the minimum version due to performance reasons is a Model 2. The MagicMirror-software is written in javascript and bundled via the Electron-app. While this makes it easier for me to hack on it myself, it takes it toll on the processing power of the PI.
But that isn't a crisis but merely an opportunity to buy a brandnew Model 3. So a few days later I have a package in the mailbox and the installation process works flawlessly. In the meantime I started to assemble...

Day 101 - A trip to the hardware store

Okay, so first I get rid of all the plastic covers of the display. What do I need now...

  • Simple image frame just large enough for the display to fit into
  • Some semi-transparent, half-reflecting sheet to put on the frame glass
  • A HDMI-DVI converter since the display is so old that it only has VGA or DVI inputs
  • A VESA wall mount
  • Screws and cables

A quick trip to the local hardware store yields a nearly perfectly fitting frame. The sheet is ordered on the internet. While I wait for it to arrive, I already start to assemble the display into the frame.

Day 102 - Do it yourself

The semi-transparent sheet arrives but damn, attaching it to the glas is mighty difficult and doesn't yield the best results if you are living together with cats :-) Put it on my list of "Things to later improve": Buy something finished from the shop. Like the paintwork of my g4pc a craftsman can do it a thousand times better in his shop than me in my home.

Day 103 - Don't move!

Time to check how expensive this thing will be if I run it 24h nonstop for a year.... Lets see, the pi takes around 4W, the monitor 25W, round it up to 30, multiply it with a lot of zeros to get the per year amount. Multiply by 0.28cents per kwh... 70 EUR??? Nope, too much, the monitor has to be put in standby if noone is near...

Good thing there is a module for that, the motiondetector. The documentation is non-existing but the biggest problem: What camera should I use? Should I buy one especially for the pi and which one is working out of the box with this software? Let's google for it... Wait, what... the ps3 move camera is working with the pi too and can be plugged just into the usb ports? Good thing I have one of these lying around too, this sounds way to good to be true (and I didnt play those ps3 move games anyway :-)

Aaaaand it works right out of the box! Well, if you are not using an old linux kernel where support for it was temporarily broken. But with the latest OS and drivers it's not a problem.

As for the motiondetector, since it is open source, not very well documented and I had the urge to add a threshold to the detection, I forked it and made my own module using this config:

config: {
            captureIntervalTime: 1000,
            scoreThreshold: 100,
            timeout: 120000

So now the monitor gets suspended if no motion is detected for a minute. That puts the energy consumption down to less than 1W for the display, leaving a total at 5W. Now the costs of running the mirror are at a reasonable level.

Day 104 - Hanging around

So this is how it currently loks like:

Days to come - Listen to me!

With the ps3 doing the motion detection, I now turn my eye to voice recognition. How cool would it be if I could tell my magicmirror to turn on the music? So, lets see if the ps3 microphone is also working outofthebox. Plug it in aaaand...

Well, it does show up using this command:

cat /proc/asound/cards
            0 [ALSA]: bcm2835 - bcm2835 ALSA
                bcm2835 ALSA
            1 [CameraB409241]: USB-Audio - USB Camera-B4.09.24.1
                OmniVision Technologies, Inc. USB Camera-B4.09.24.1 at usb-3f980000.usb-1.3, hi

but how do I find out if it actually records? NielsMayer has some good tips on how to get your mic configured and also wrote this command:

arecord -vv /dev/null -r 16000 -f S16_LE -c 4 -D iec958:CARD=CameraB409241,DEV=0 /dev/null < /dev/null
            Aufnahme: WAVE '/dev/null' : Signed 16 bit Little Endian, Rate: 16000 Hz, 4 Kanäle
            Hardware PCM card 1 'USB Camera-B4.09.24.1' device 0 subdevice 0
            Its setup is:
            stream       : CAPTURE
            access       : RW_INTERLEAVED
            format       : S16_LE
            subformat    : STD
            channels     : 4
            rate         : 16000
            exact rate   : 16000 (16000/1)
            msbits       : 16
            buffer_size  : 8000
            period_size  : 2000
            period_time  : 125000
            tstamp_mode  : NONE
            period_step  : 1
            avail_min    : 2000
            period_event : 0
            start_threshold  : 1
            stop_threshold   : 8000
            silence_threshold: 0
            silence_size : 0
            boundary     : 2097152000
            appl_ptr     : 0
            hw_ptr       : 0
            #+                                                 | 01%

See that 01% at the end? Yeah, background noise that gets detected! Awesome! With this out of the way the next step is to install some modules for the mirror to read and process the voice commands...

To be continued...