CMUSphinx4 SR

Speech recognizer based on the CMUSphinx project. Additional language models can be downloaded from Sourceforge and Voxforge. You can also learn your own dictionary and language model and reuse the standard English acoustic model. To automatically create a dictionary and language model file from smaller text files simply go to CMU lmtool page. For more details see the CMU Building Language Model tutorial. Once created you can use your dictionary and language model with an existing acoustic model e.g. en-us.

Notice: If you get bad results check that the sampling rate of your audio files match the one used to train the language model (see CMUSphinx FAQ). The included models where trained with a sampling rate of 16kHz.


Audio Column
The audio column to process
Fail on error
The node fails when an error occurs during processing. If input has multiple AudioFiles, setting "Fail on Error" results in the Node failing when Speech Recognition of one one more AudioFiles fails. If not set, the Node will complete with missing values for all SR columns of the failed AudioFiles.
Audio Model
Build in audio models. Please notice that the included models where trained with a sampling rate of 16kHz.
Selected acoustic model directory
The acoustic model to use. Supports KNIME protocol e.g. knime://knime.workflow/data/acoustic
Selected dictionary file
The dictionary file to use. Supports KNIME protocol e.g. knime://knime.workflow/data/knime.dict
Selected language model file
The language model to use. Supports KNIME protocol e.g. knime://knime.workflow/data/knime.lm

Input Ports

Table with audio column

Output Ports

Audio table with recognition result


This node has no views




You want to see the source code for this node? Click the following button and we’ll use our super-powers to find it for you.