The ATR design is optimal design for speech, where the endpoint of the call (the phone, device, or kiosk hardware) communicates feedback to the server about the quality of the connection and background noise.

Distributed Speech Recognition is good Speech design.
DSR hardware (on client) dramatically boosts results (see below). Delivering access to electronic information to users without a PC widens and deepens your demographics for the content you are delivering. Even as speech continues to perform well on legacy platforms, embedding systems multiply accessibility and performance. As cell phones get smarter ATR gets better

Speech recognition based technology scans verbal input for utterance of specific phrases or statements, compiled in a grammar (a directed dialogue) and executes commands based on those statements. This technology is known as Automated Speech Recognition or ASR. When you can apply mPath or mPathy, you have distributed speech, which dramatically improves your results. Below is an illustration that describes the basic process note ‘mPath’ refers to the collection of multiple microphone inputs and their gate information ‘mPathy’ refers to the process applied against those inputs.