PHONEMIC RECOGNITION METHOD FOR CONTINUOUS SPEECH

摘要

The phonemic recognition device for continuous speech comprises an analyzer, a table value memory block of elementary similarity measures and speech recognized signal memory block, an integral similarity measure calculator and a controller. In the device there is a teaching selection memory block, a a cluster analysis processor, a phonemic parameter memory block, an orthographic text and phonemic transcription memory block, a vector quantizer where the analyzer output through a speech signal memory block to a vector quantizer input and through the teaching selection memory block to cluster analysis processor input is connected, the outputs being correspondingly connected to a vector quantizer input the outputs being correspondingly connected to a vector quantizer input, the corresponding inputs being connected to orthographic text memory block and phonemic transcription outputs and a vector quantizer output is also connected to table value memory block input of elementary similarity measures, the outputs being connected to integral similarity measure calculator inputs. The corresponding orthographic text memory block and phonemic transcription outputs are connected to a corresponding table value memory block input of elementary similarity measures and corresponding table value memory block of elementary similarity measures and corresponding and an integral similarity measure calculator input and controller outputs are connected to corresponding device unit inputs. .