From HemeraBook
Revision as of 18:09, 29 November 2011 by Projettwk (Talk | contribs) (Speech core module)

Jump to: navigation, search


configure: error: no acceptable C compiler found in $PATH

You must prepare your computer for compiling source code.

make: *** No targets specified and no makefile found. Stop.

First, you must prepare your computer for compiling source code to ensure having GNU version of make.

Then, ensure a Makefile is available in the directory in which you launched the command.
The most of third-party tools uses autotools, so you usually need to launch ./configure to create such Makefile.

There is WARNING or ERROR when using Hemera

In this case, follow these instructions until you find a solution:

  • fix the issue if the message gives instruction
  • check the log (indicated on standard output) for more details
  • lookup solution on other section of this Troubleshooting page
  • ask for solution on the Help Forum

GNU version of which not found. Please install it

Hemera requires GNU version of which, which is natively provided by some GNU/Linux distribution.
Anyway, it is not the case of Debian-based distributions, in which case you must install it from source.

You must update your LANG environment variable to use the UTF-8 charmaps ('XXX' detected)

Hemera requires UTF-8, update your OS to use UTF-8.

Although the current OS locale 'XXX' defines to use the UTF-8 charmaps, it is not available (checked with 'locale -a')

Hemera requires UTF-8, update your OS to use UTF-8.
This locale must be installed on your OS, which can be confirmed using the locale command.

lt-sox FAIL sox: Not enough input filenames specified; and SoX usage (lots of lines) is shown when Hemera attempts to play/rec sound

You may have installed SoX from source code.
With this version, input and output must be specified on command line. For Default Sound Device (either as input or output), the option -d is use.
Update your configuration file comparing with the template in which corresponding option has been added later.

lt-sox FAIL formats: can't open output file `/dev/dsp': Permission denied

You may have installed SoX from source code.
It means your version of SoX version is missing support to your Sound Device. You must ensure having support to it (usually Alsa or PulseAudio) and compile source code again.
Ensure that end of report given by configure indicates support to your Sound Device.
For instance:

 alsa (Linux)...............yes

Unable to record sound with microphone

You must ensure you can record with your microphone.
You can then perform sound check.

Speech core module

Detected version XXX of espeak which needs --pho option

This error message can be seen since version 1.44.0 of espeak, which needs a new option to be used as Phoneme mnemonics writer.
To fix this issue, check/update the following configuration element


There was error while using mbrola, install corresponding file or create symbolic link

This can happen if some mbrola voices model are not installed or reachable.
To fix this issue, adapt these instructions.

/usr/share/mbrola/voices/fr1 not found

Since version 1.44, espeak changes the way it looks up for language files.

Follow these instructions according to situation.

Directory /usr/share/mbrola/voices exists

On Debian-based distributions, this directory is usually created by mbrola language packages (e.g. mbrola-fr4); so to fix this issue, you can create a symbolic link:

ln -s /usr/share/mbrola/voices/fr4 /usr/share/mbrola/voices/fr1

Directory /usr/share/mbrola/voices does NOT exist

This issue has been seen on recent version of Redhat-based distributions (e.g. Fedora 14).
To fix it, perform these instructions as root:

mkdir -p /usr/share/mbrola
ln -s HEMERA_TP_PATH/speech/data/language /usr/share/mbrola/voices
ln -s /usr/share/mbrola/voices/fr4 /usr/share/mbrola/voices/fr1

Speech Recognition core module

lia_phon is stuck on "Initialisation des tableaux 1.2.3-grams"

If you have an x86_64 OS architecture, you must force lia_phon compilation to x86.
You can use the patch provided by the Hemera little speech recognition model creation project .

sphinx3_decode failed in Segmentation Fault

There is some incompatibility between x86_64 old version of sphinx3_decode and UTF-8 language model.
You may follow these instructions.

there is no "answer" when I speak in microphone

You must ensure you can record with your microphone.
You can then perform sound check.