Floating Point Numbers on 16 Bit Processors and Digital Number Systems for DSP

I came across an interesting paper in the July edition of IEEE Communications Magazine.  The article describes some of the challenges of doing DSP style calculations with number systems [1].  I didn't take time to fully understand everything in the paper, but I understand it from a high level and wanted to make a note of this article for future reference.  I think the main point of the article was to describe a new floating point number system that is efficiently implemented on 16 bit processors.  I'm not sure of the value of this based on my experience; often 32 bit processors are cheap enough and highly competitive price wise with 16 bit processors now that 16 bit processors are rarely chosen.

[1] Richey, Manuel, and Hossein Saiedian. July 2009. "A New Class of Floating-Point Data Formats with Applications to 16-Bit Digital-Signal Processing Systems." IEEE Communications Magazine. Vol. 47, No. 4.