Frequently Asked Questions

LumenVox Speech Recognition on Asterisk

How can I move my licenses?

Licenses are tied to specific machines. When you install the License Server, you will generate a file called info.bts that describes the hardware in your system. You then upload this file to our Web site, and we generate a license file that is based on that hardware configuration. The license file will not work on a different machine. However, if you format the machine, the license will still work on the machine provided the hardware configuration has not changed.

We encourage users to keep their licenses on a single machine. Speech Engine clients loaded on different machines can query the central license server for licenses, and you can freely change the machines on which the Speech Engine runs.

We understand that users do sometimes have a need to move licenses onto new hardware. If for some reason you need to move a License Server, Contact Us and briefly explain your situation and our support team will work with you to address the issue.

How do I download or install my product?

When you purchased a Speech Engine license through Digium, you should have received an email with instructions on download and installation. If you do not have that email message, Contact Us for assistance.

What files do I need to download?

You will need to download a Speech Engine client and a License Server. If you are using the Open Source edition of Asterisk, you will also need the Asterisk connector for your version of Asterisk (Asterisk Business Edition does not require the separate connector).

Why won't the software install?

Make sure you are following the instructions carefully, especially if your Linux experience is limited. The most common reasons the LumenVox Speech Engine or License Server fail are:

• The user did not log out and log back in after installing the products (the products require that some environment variables be set).
• The user has mistyped a command. Remember that on Linux, capitalization matters, so check upper and lower case letters.
• The account under which it was installed did not have administrative privileges. You cannot install the Speech Engine or License Server as a regular user; you must be logged in as root or as another user with administrative access.

Which Linux distributions does LumenVox support?

We support recent releases of the following Linux distributions:

• Red Hat Enterprise Server 6, CentOS 6 (64-bit only)
• Red Hat Enterprise Server 7, CentOS 7 (32 & 64-bit)

Our software may work on other distributions, but we have not tested it and cannot provide any assistance with getting it to run on those distributions.

We can support new distributions if there is a significant demand for them, but it takes us a considerable amount of time to certify our installation on any given distribution. We are willing to entertain requests from high–volume customers who need a specific Linux distribution supported. Contact Us with specific inquiries.

Are any example programs for speech recognition on Asterisk?

Yes. See our Pizza Demo for example code that uses the Dial Plan to create a speech recognition program on Asterisk.

If you just wish to make sure that the Engine is installed properly, a test program is included in the Engine installation directory, which is /opt/lumenvox/engine_7.0/examples/ by default. Navigate to that directory, type make and press enter. Then type ./example 127.0.0.1 and press enter. If the Engine is installed correctly, you will get a looping decode from the Engine that will give you an interpretation of "8587070707."

Why is my speech application not producing the sort of accuracy I expect?

If you are using a simple grammar, such as the built–in yes/no grammar, there is likely a hardware or configuration issue if you are getting low accuracy. Open the lumenvox.conf file, located in /etc/asterisk/ by default. There are a number of parameters there that you should consider changing, especially those that relate to detecting barge–in and end–of–speech. A common problem is that the Speech Engine is not parsing the entire utterance because one of these settings is incorrect given the environment or speaker.

Troubleshooting speech applications is a process called "tuning." Applications with larger grammars are generally more difficult to tune. For instance, in a speech recognition auto attendant or call router, you may have two or more names that sound very similar. This is going to be difficult for the Engine to resolve. Consider adding weights within the grammar to one name if you know it will be asked for more often than another.

LumenVox has a Speech Tuner available for troubleshooting these sorts of problems. A separate, annual Speech Tuner subscription license is required. Contact Us for information about purchasing a Speech Tuner license.

There are lots of tricks in developing grammars and applications that you can use to achieve higher accuracy. If you are new to developing speech recognition programs, you may be interested in taking one of our training classes, tailored specifically to your needs.

Also consider reviewing the section in the Speech Engine help files about working with SRGS grammars. For words that are commonly mispronounced, you may wish to add extra pronunciations using phonetic spelling; this is also documented in the help files.

What does the error "version 'GLIBCXX_3.4.6' not found" mean?

The following error message:

/usr/lib/libstdc++.so.6: version 'GLIBCXX_3.4.6' not found

Means that the glibc libraries on your machine are probably out of date. Update your machine's libraries and try running the application again.

What does the error "cannot restore segment prot after reloc" mean?

This is an error related to Security Enhanced Linux. Disable SE Linux (go to System Settings and choose Security Level, go to the SE Linux tab, and uncheck the "Enable" button), reboot, and try reinstalling the product.