Semantics of Simband data

At its core, Simband is a platform for collecting and analyzing biophysical data for preventative health. As such, data on this platform has certain traits: it is, at most, time-coded, continuous and numeric. This is why the Simband API describes data as a stream.

Before diving into quantitative details of the stream and Simband Runtime, let's look at Simband's system block diagram for an overview.

  • The biophysical data comes from the heart of Simband: the sensor module. This documentation covers our reference implementation, Simsense.
  • The data gets signal conditioned and processed by M0 + M4F.
  • The data is provided to Simband Runtime, which categorizes it into streams and makes it accessible on demand.

Let's find out how streams are created and used by Simband.

What is a stream?

A stream is a single continuous sequence of time-coded numeric data. Any numeric data on Simband can be expressed as a stream, which exists in parallel with streams produced by other sensors. By definition, any stream may have only one numeric value at any given point in time, or no value at all. All of the stream data on Simband is float-32.

For example, each PPG channel is a stream, but an accelerometer is divided into 3 streams (x, y, z). This is because each axis may have a single numeric value at any given point in time. Learn about types of streams.

The flow of data

The primary source of data in Simband is Simsense. Each sensor in Simsense generates data at a pace dependent on its speed of operation, events and context. Some sensors generate data each second, while others might be inactive for minutes at a time.

All the data, when generated from the sensor, is marked by a timestamp. Depending on the mode of operation, sensor data is either buffered locally in Simsense or data is sent to Simband using our communication protocol. Data generated from different sensors are normalized to the same clock and to a 128 Hz sampling rate before being sent to Simband.

This diagram shows how data flows from the sensor module to Simband and how data is made accessible to various algorithms, the Simband UI and ARTIK Cloud.

The Simband component simsensed receives the data from the sensor module, normalizes and categorizes the data, and sends the data to the datad.

simsensed categorizes data into its stream type and normalizes the timestamps to UNIX timestamps.

datad stores the data into the database and makes the data available when requested by an algorithm or the Simband UI. datad is also connected to ARTIK Cloud and uploads the stored data based on the availability of WiFi. In an ideal network environment, you will be able to see data flowing in and out of ARTIK Cloud with a delay of a couple of seconds (this is mostly due to network latency).

Where algorithms come in

Algorithms running on Simband elaborate the data, store the results locally and push them to ARTIK Cloud, and also display them to the user. An algorithm may request the current value in the stream coming from a given sensor, or it may read all values in a given time frame, and Simband will provide the correct information based on the availability of the data.

Streams are immutable—i.e., algorithms cannot modify a stream of data linked to a sensor. When an algorithm wants to store data to be used later, it will effectively create a new stream of data fed by the algorithm itself.

The power of this implementation is that each algorithm is the "owner if its fate," and can alter it at any time in the way it wants. Other algorithms, meanwhile, can also "dip" into that stream and benefit from the computations that happen in Simband without repeating the same work.