viennajack
02-23-2003, 11:35 PM
I've been experimenting with the Clie NX70V voice recorder and Dragon NaturallySpeaking version 5.0, using Dragon NaturallySpeaking to transcribe the recorded voice audio from the Clie. Dragon NaturallySpeaking supports transcription of recorded .WAV files from mobile recorder devices. It is directly compatible with Dragon's NaturallyMobile recorder, and the Sony Memory Stick IC Recorder ICD-MS1. It is not directly compatible with audio files recorded with the Clie.
However, with a simple conversion it is possible to transcribe Clie .WAV files directly into text, with reasonable accuracy.
First, it is necessary to train Dragon NaturallySpeaking to how your voice sounds when played from the Clie .WAV file. This involves reading approximately 20 minutes of sample text provided by Dragon NaturallySpeaking into the Clie recorder, converting the Clie .WAV file to the format that Dragon NaturallySpeaking uses, and telling Dragon NaturallySpeaking to "train" to your recorded voice using this 20 minutes of recorded audio. Once Dragon NaturallySpeaking is trained to your voice, it will be able to transcribe your recorded memos from the Clie. Accuracy can be improved by recording additional training texts on the Clie and allowing Dragon NaturallySpeaking to train on these additional texts. Dragon NaturallySpeaking provides several training texts that can be used for training to your recorded voice.
The key to making this work is the file conversion process. Clie WAV files are recorded using a 4 bit, 22 kHz, mono ADPCM format. Dragon NaturallySpeaking expects transcription .WAV files in 16-bit, 11.025 kHz, mono PCM format.
Windows includes a utility called Sound Recorder. It can be found under the Start Menu, under Programs > Accessories > Entertainment. To convert a WAV file, copy it from the Memory Stick (or Clie internal memory) to the PC, then open it with Sound Recorder. Go to the File menu, select Save As, and under Attributes, choose "11.025 kHz, 16-bit, mono, 21 kb/sec". Save the file using this format and it will be compatible with Dragon NaturallySpeaking.
I haven't used this to dictate actual letters or documents yet, but I have tested it by reading from catalogs and books. Accuracy is not bad! Let me know if you figure out a useful application for this.
Have fun!
Jack
However, with a simple conversion it is possible to transcribe Clie .WAV files directly into text, with reasonable accuracy.
First, it is necessary to train Dragon NaturallySpeaking to how your voice sounds when played from the Clie .WAV file. This involves reading approximately 20 minutes of sample text provided by Dragon NaturallySpeaking into the Clie recorder, converting the Clie .WAV file to the format that Dragon NaturallySpeaking uses, and telling Dragon NaturallySpeaking to "train" to your recorded voice using this 20 minutes of recorded audio. Once Dragon NaturallySpeaking is trained to your voice, it will be able to transcribe your recorded memos from the Clie. Accuracy can be improved by recording additional training texts on the Clie and allowing Dragon NaturallySpeaking to train on these additional texts. Dragon NaturallySpeaking provides several training texts that can be used for training to your recorded voice.
The key to making this work is the file conversion process. Clie WAV files are recorded using a 4 bit, 22 kHz, mono ADPCM format. Dragon NaturallySpeaking expects transcription .WAV files in 16-bit, 11.025 kHz, mono PCM format.
Windows includes a utility called Sound Recorder. It can be found under the Start Menu, under Programs > Accessories > Entertainment. To convert a WAV file, copy it from the Memory Stick (or Clie internal memory) to the PC, then open it with Sound Recorder. Go to the File menu, select Save As, and under Attributes, choose "11.025 kHz, 16-bit, mono, 21 kb/sec". Save the file using this format and it will be compatible with Dragon NaturallySpeaking.
I haven't used this to dictate actual letters or documents yet, but I have tested it by reading from catalogs and books. Accuracy is not bad! Let me know if you figure out a useful application for this.
Have fun!
Jack