Can a human speak to a machine in a machine’s language? Yes, with The Chirp Human SDK.

Chirp’s audio technology uses sound to transfer data from one device to
another. The sender plays a series of tones representing the data, and a
receiver can then decode any data it hears. Think of it as a message between
two devices.

It is a simple concept which enables billions of devices to talk to one
another without the need for expensive or complicated hardware. The audible
version of our technology plays a series of tones to broadcast a message to
other devices. The sound that is played is the message, meaning that what
comes out of a device’s loudspeaker can be heard both by humans and other
machines — and this audio can then be decoded by the receiver to get the
original message.

How simple is it to send and receive a Chirp signal? We set out on a fun
mission this week — to see if we could use our own voices to send a message to
a device using Chirp. The result — a new dimension to the phrase human-to-
machine interaction.

To understand how this works, let’s first look at how the data is sent.

Transcribing a chirp

Below is a representation of a single recorded Chirp message. It shows a
sequence of different frequencies — the whole sequence is played in around 2
seconds in our standard product.

This representation is a little abstract, but we can see that chirps are
simply sequences of different ‘notes’. Of course, there is a much more human-
readable method of representing a collection of notes: musical notation.

Let’s see how this exact same series of pitches looks in musical notation:

Before going further, let’s double-check that the notes above map onto the
frequencies that the receiving device is expecting:

In order to simplify the playing of these notes, we decided to change the
frequency ranges that the devices were expecting, in order to bring them into
a more manageable range for us to play, sing or whistle…. And then we tried to
send a message to a Chirp-enabled receiving device…

Here’s Anya, our head of marketing, having some fun by sending a message to
her tablet by simply whistling to it.

The interesting thing here is that the device is not using voice recognition
to understand us. This is a human speaking the machine’s language.

Teach your machines to sing

We pride ourselves on the way that our data-over-sound technology is uniquely
easy to understand, and is a transparent and honest form of machine
communication that is open to inspection by a human listener. If you’d like to
explore teaching your machines (or humans) to sing with Chirp, get started at
our Developers Hub, where you can find:

  • Cross-platform SDKs (iOS, Android, Python, C, JavaScript, Raspberry Pi, Arm Cortex, and more — get in touch if you can’t see what you want as we likely have some alpha code that you can play with)
  • Example apps and code snippets
  • Easy setup and integration — they can be integrated within your app, product or project within a matter of minutes (here’s a video of us doing it in a new project in under 4 minutes)
  • Support - we are always here to help.

Or if you’d like to talk further, contact us directly at contact@chirp.io