- Installation
- Administration
- Programmer's Guide
- Grammars
- MRCP Server
- FAQs
The LumenVox Speech Engine features several configurations in its handling of noise reduction.
Noise reduction is accomplished by taking a sample of the background noise in the audio being streamed to the Engine. As long as it’s done at the beginning of the stream – before the caller starts speaking – the Engine can gauge how much of the noise after the caller has started speaking is just background noise. This computation takes less than a second.
This technology is particularly useful in noisy environments, such as a crowded place, or in a car.
There are now 4 settings to use to counter noisy backgrounds. They are configured in the NOISE_REDUCTION_ENABLE parameter of your client_property.conf file.
NONE
Description: Disables all noise reduction. We do not recommend disabling all noise reduction, as the increased presence of background noise will also likely increase the time needed to decode speech. However, this may be used in debugging another system�s noise reduction features, or examining the audio itself.
Value: 0
DEFAULT
Description: This is our default Noise Reduction algorithm, custom built for our Speech Engine and recommended for most of its applications. It is automatically activated when you install the Engine.
Value: 1
ALTERNATE
Description: This is a new Noise Reduction model for the LumenVox Speech Engine. It works by filtering different types of noise than the Default model. There are no criteria for which one will work best in a given environment, but it is recommended that you start with the Default model.
Value: 2
ADAPTIVE
Description: This is another new model. It adapts to the noise signal, and continuously updates its noise estimation. This is particularly well suited to settings where the background noise will fluctuate, like the outdoors or a car. However, this model may not function as well in stationary noise environments, like a home or an office.
Value: 3
In order to change your Noise Reduction model, you will need to modify your client_property.conf, which is in /etc/lumenvox/ for Linux users, and C:\Program Files\LumenVox\Engine\ on Windows.