Accessible checkboxes usually require two operations to control: first the user needs to 'focus' the checkbox (usually by using the 'tab' key to move through the interface until they reach the right control) and then they need to change the state of the checkbox (usually using the 'space' key). Sighted users will skip through the page using visual cues such as headings or images but a visually impaired user will tab through the page from link-to-link or control-to-control until they reach the one they want without reading any text in between.
That's your first hurdle: how does the user know what the control is or does? - Your control will need to be well labelled so that the user knows exactly what the control does without relying on any text, icons or images around it. You might also want to look into 'ARIA roles' to help with marking interactive elements in a more accessible way.
Looking at the code you're proposing to use, I can also see another issue: How does the user know that a change has taken place once they have operated the control? In the code you linked to, the only non-visual change is that the checkbox reports as 'checked' when activated. I don't think that alone is enough to signal the full extent of the control's purpose and status.