chatchavan / moov

Functions for reading data out of Moov sensors

Moov Swimming Data Toolkit

This R package provides functions for reading data out of MOOV sensors. Currently supports only data from swimming sessions.

Retrieving Moov data

To use this package, you first need to retrieve data from Moov that is stored in your device. We will need the file user.db and the folder swimming. Both of them are stored in /Applications/moov.core/Library/Application Support/user_data/( user id )/.

iOS: Use iExplorer (the free version is OK) to browse your iOS device backup.

Android: See the instruction in Marin's code.

Usage

Install and load the package:

devtools::install_github("chatchavan/Moov")
library(Moov)

See an example here

Structure of Moov data

The user.db file is an SQLite database containing summary data for each sesssion. There are sub-folder inside the swimming\(some number)\ for each session. Each of these folders contains three files:

• swimming_result.pb: A binary file containing sensor data from that session. (More on this below.)
• swimming_post_analysis_result.pb: A binary file. Probably contain summary of that session , e.g., the number of strokes in each lap. I haven't managed to parse this.
• swimming_result.pb.uploaded: always zero bytes. I suspect this is a marker file indicating that the data is uploaded to the server

swimming_result.pb

Most of this file is the sensor data. Before you can get to the data, you need to extract them from the header and trim out the tail. The actual sensor data seems to be prefixed by one of the following byte sequences:

older files: 0900 0000 0000 0000 0011 FFFF B0B4 48FC 0341 9A01
newer files: 0108 0C10 6018 0320 0A2A

Each frame of the sensor data seems to be 2 bytes. I read them as integer and plot the values over the number of frame. The periodical structure seems to mimic swimming pattern. (Strokes and laps are visible in the plot.)

The end of each file is tailed with a bunch of FFFF bytes.

user.db

Here's an example record from user.db:

- id: "1"
- session_id: "swimming-xj9p6a2uzi8u010jmwpjyc2"
- user_id: "6234565289443328"
- title: "Afternoon Swim"
- workout_type: "2"
- program: "0"
- calories: "442.163146972656"
- location_text: ""
- location_long: "0.0"
- location_lat: "0.0"
- start_time: "1461945492"
- end_time: "1461948928"
- duration: "3394"
- magnitude: "1900.0"
- summary: ""
- highlight_type: "0"
- highlight_text: ""
- status: "1"
- upload_status: "3"
- created_at: "1465396302"
- updated_at: "1495716980"
- server_updated_at: "1495713919"
- env_info: "{""app_build"": ""3.11.910.160"", ""app_name"": ""moov.core"", ""app_version"": ""160"", ""devices"": [], ""gps_enabled"": true, ""phone_model"": ""iPhone8,1"", ""phone_os_version"": ""9.3.1""}"
- program_specific_data: "{""lap_count"": 38, ""lap_length_in_lap_unit"": 50, ""lap_unit"": 0, ""stroke_count"": 905, ""user_height"": 1.7599999904632568, ""user_weight"": 69}"
- local_cache: "{""distance"": 1950, ""distance_per_stroke"": 2.1546962261199951, ""lap_length"": 50, ""stroke_rate"": 2.018596887588501, ""version"": 3}"
	- Key names are changed from record 115 onward
	
	- This field is NULL for the last 6 records
	- The "distance_per_stroke" doesn't match with "magnitude / stroke_count"

• workout_type: always "2" for swimming.
• calories: in kCal
• duration: Swimming duration in seconds. This matches what shown on the MOOV summary screen This is not necessarily less than "end_time - start_time"
• magnitude: The total distance in meters
• local_cache: JSON data with some performance metric calculation
  • This field is NA from recent records
  • The new version of the file, the name of some fields are changed:
    • distance_per_stroke → main_distance_per_stroke
      • stroke_rate → main_stroke_rate

Tested sensors

• Moov Motion Sensor (Moov Classic)

moov's Issues

newer filetypes?

I would love access to the pure sensor data of the Moov, using the newer Moov Now.
Im definitely quite limited in my knowledge with R, and binary filetypes, so im kind of stuck getting this to work with the files generated by the moov now.

I adjusted your header_new <- c(2049L, 4108L,6240L,8195L,10762L)

And can now see quite elaborate data structures, with what seems to be 9 different data sets? corresponding to the 9 sensors in the moov i would guess. (Accel, gyro, magneto)
But seems to me something is not right here, and the data is not decoded correctly...