SensorTile.box #FromTheCustomer: Fridge opening detection (Mode2 + Cloud)

[Eleon BORLINI]

Do you know how to monitor your fridge and detect whether a family member is crawling into the kitchen in the middle of the night and indulging in a bit of snacking? Follow this Share you activity post to get an insight on a practical use of the SensorTile.box!

[Pgarg.11]

can you tell me which functions should i use so that the data capture will be same for the differnt sensors i am also using the sensors Input Sensors:

1.      Humidity (HTS221)

2.      Temp (STTS751)

3.      Pressure (LPS22HH)

4.      Vibrometer (LSM6DOX)

5.      Gyroscope (LSM6DOX)

6.      Microphone(MP23ABS1)

cuz in the output of csv file that is uploaded in attachment, i am getting the date and time different as when accelerometer starts logging at the same time gyro stop logging (see the A column A of "microphone and accelerometer and gyro) but i want to log the this column A (date) should log same time for all the sensors.

[Eleon BORLINI]

Hi @Pgarg.1​ , not sure if I catch your question... no "Function" is needed, because the sensors outputs data at different ODR but they are synchronized in the csv file, meaning that if acc outputs data at 50Hz and gyro at 52Hz, you will have the same acc data once every 50 samples more or less. Attached the overall data log.

Regards

[Pgarg.11]

@Eleon BORLINI​ absolutely, you caught the issue i also want the integrated output as you only one csv file unlike i was getting 6 csv file for all different 6 sensors.

but aslo in the output.csv the log of microphone is not reflected

"The app you designed(prova log), there you used two inputs more (RTC (date and time)) as showed in above screenshot "

Firstly, tell me are the RTC sensors you used for synchronization (if yes then why the date in your csv file "output.csv" is not changing after 24 hours) ??

Secondly, also i am using the same input sensors but my app is not uploading on the st.box (can be seen in the video attached) and,

1. for the input sensors ODR would be kept at "default frequency" ( as i want the synchronized and integrated output for the sensors as you have showed in the file but also for microphone too)
2. should i use the available functions (actually what role they are playing can ou tell me)
3. and where should i stream the output (i guess you are using the sd card so i did the same).

[Eleon BORLINI]

Hi @Pgarg.1​ ,

>> but aslo in the output.csv the log of microphone is not reflected

This is right, the microphone has to be treated separately from the other sensors, by now. The app mixes the .csv data and the .wav data resulting in a wrong .csv file. Better to exclude the microphone by now.

>> tell me are the RTC sensors you used for synchronization

The highest ODR drives the timestamp, e.g. the ODR 52Hz of gyro or acc. So, for the sensors with lower ODR, sometime you will see the same data reported.

>> Secondly, also i am using the same input sensors but my app is not uploading on the st.box

Right, you get into an error, but if you then connect to your ST.box you will see that the app has been correctly downloaded on ST.box MCU. We have to solve this pop-up error after separating mic raw data (.wav) from other sensors (.csv)

Regards

[Pgarg.11]

ok now after capture the data logs now how can i  Clean, label Data and Build ANN topology??

[Eleon BORLINI]

Hi @Pgarg.1​ , if you want to deploy the MLC feature of the inertial module DSOX on the ST.box, after the dataset creation you have to create an .ucf file with Unico and then upload it on the DSOX through the ST BLE app: for the description of this steps you can refer to this link for the MLC on Unico tool and to this other link for the ST BLE implementation. On the other hand, if you want to implement a more complex NN than a decision tree, you can start from the FP-AI-SENSING1 function pack that can run on the STM32L4 of the ST.box and create your code. Regards

[Pgarg.11]

@Eleon BORLINI​ yes, i want to implement more complex NN.

but for that as you have mentioned in FP-AI-SENSING1 (for AI Data Logging in UM2524 page number 31 to 36, because i want that type of labelled data like you defined label for walking and running) that we have to first flash sensing.bin "STM32CubeFunctionPack_SENSING1_V4.0.1\Projects\STM32L4R9ZI-SensorTile.box\Applications"

in order to get the Clean, label Data for Building of the ANN topology but if you remember i asked you in the private chat about that after flashing the binary, i am not able to get the logged data on onboard sd card so that i can build ANN topology and by the way how to TRAIN NN MODEL (using which Deep Learning Frameworks).

Do i have to flash the binary using ST LInk or Nucleo or can i flash directly using the ST BLE Sensor using my smartphone(because i have tried using smartphone that's why i am saying i am unable to collect the data log on sd card) ?

and please tell me the steps to sync the .wav file with .csv file because i must have to use the microphone, i can't leave the microphone?