A low cost dynamic vocabulary speech recognizer on a GPP-DSP system

Continuous speech recognition is a resource-intensive algorithm. Commercial dictation software requires more than 10 Mbytes to install on the disk and 32 Mbytes RAM to run the application. Because of the resource requirement, such a system can not be implemented in a low cost and low power embedded system. We propose a design of dynamic vocabulary speech recognizer that will fit in a DSP-GPP (general purpose processor) architecture. The computation intensive, small footprint recognizer engine runs on the DSP; and the computation non-intensive, larger footprint grammar, dictionary, and model acoustic components resides on the GPP. The recognition models are prepared on the GPP and transferred to the DSP, the interaction among the application, model generation, and recognition modules is minimal. The result is a speech recognition server implemented in a low cost embedded system. The application can dynamically create flexible vocabulary to suit different recognition contexts. It still does not do large vocabulary dictation; however, it provides unlimited recognition contexts with unlimited vocabulary, all these implementable in a low cost embedded system