SignalML provides a simple and effective way of encoding the metainformation needed for a proper interpretation of digital time series, stored in different formats. Unlike the actual software (programs), created in thousands for conversions, display or analysis of data in particular dataformats, SignalML encoding requires only one instance (metainformation SignalML file) for a given format to make it readable by any compliant software. E.g. if a hardware producer wants to make the data, stored by his signal acquisition systems, accesible by a (future) wealth of software of other providers (including Open Source), he does not need to supply low-level input/output/conversion routines for any imaginable programming language and operating system. All it takes are up to a few dozen lines of standard XML.
This approach greatly simplifies also the task of writing software capable of accessing data in more than one format. Classically, programmers needed to write separate low-level routines for each different format or format feature. After a software project was closed, it was extremely difficult to add support for another dataformat. With SignalML we can write just one routine for reading any dataformat based upon its meta-description in SignalML.
Finally, for any interested scientist or clinician, writing such a meta-description for most of the formats is simpler than programming low level I/O routines, and opens access to the wealth of compliant software. The availability of the multiplatform browser/annotator, mentioned in section 3.3, should be seens as the first step in creation of this ``wealth'', since such an open and user-friendly program was, at least in the field of EEG, needed and missing for years. But above all we hope that this approach will be accepted by the community. This would expectably lead to development and/or adaptation of wide variety of compliant software. Adopting any existing software to use SignalML is a task comparable to writing just one more I/O routine, since standard XML parsers are readily available for most of the programming languages.
The presented idea was exemplified on the electroencephalographic and polysomnographic recordings, but it is not limited to these. Most of the formats for biomedical time series can be efficiently described within the proposed Schema--maybe after some enhancements resulting from the open discussion. This paper, the definition of the SignalML language and the accompanying software implementation provide the critical mass needed to start a discussion and collaborative effort of interested parties, leading to the development of SignalML and compliant software. We hope it will lead to a universal, elegant and widely accepted markup language, which in due time will be submitted as a standard to the World Wide Web Consortium.