适配声学模型时运行sphinx_fe出现Segmentation fault错误求助

Let's work through this segmentation fault step by step—here are the most likely fixes based on your setup and the GDB output pointing to strlen():

• Verify the feat.params file
  The error in strlen() suggests the command is trying to process an invalid string, often from a malformed parameter file. Check:

  • Is the path en-us/feat.params correct? Make sure you're running the command from the directory where the en-us folder lives, or use an absolute path (e.g., /home/your-user/cmusphinx/en-us/feat.params).
  • Open feat.params and confirm there are no empty lines, missing values, or corrupted entries. It should contain valid feature extraction parameters like -samprate 16000 or -winlen 0.025.

• Check the arctic20.fileids file
  This file lists your WAV files, and invalid entries here can cause string-handling crashes:

  • Ensure every line in arctic20.fileids is a valid, existing path to a WAV file (no typos, no missing files, no trailing spaces).
  • Remove any empty lines or commented-out lines that aren't properly prefixed with #.

• Validate your WAV files
  sphinx_fe requires WAV files to be 16kHz, single-channel (mono), PCM-encoded. Mismatched formats can lead to low-level read errors that trigger segmentation faults:

  • Use soxi your-file.wav (install with sudo apt install sox) to check the format. If it's not 16kHz mono, convert it with:

    sox input.wav -r 16000 -c 1 output.wav
    

• Reinstall or rebuild SphinxBase
  Ubuntu 16.04's default CMUSphinx packages are outdated, which can introduce compatibility bugs. Try:

  • Uninstalling the existing packages: sudo apt purge sphinxbase libsphinxbase-dev
  • Cloning the latest SphinxBase source and compiling it manually (follow the official build instructions tailored to your system).

• Dig deeper with GDB
  To pinpoint exactly which string is causing the crash, run gdb sphinx_fe and enter these commands:

  run -argfile en-us/feat.params -samprate 16000 -c arctic20.fileids -di . -do . -ei wav -eo mfc -mswav yes
  bt
  

  The backtrace (bt) will show the exact function call that led to strlen(), which can tell you if it's a parameter file issue, WAV path issue, or library bug.

内容的提问来源于stack exchange，提问作者Alexander Wentz