Registering a Personal Voice Profile
- On November 19, 2016
- personal noise reduction
Background
In the previous post we have discusses a unique technology for tuning the ambient noise reduction to your voice and therefore enable a superb personalized noise cancellation. By this point, the power and need for this technology is clear. The question that one might ask is the “price” that needs to be paid in order to use this advanced technology. In order for this technology to work, a personal voice profile must exist. So, how can this profile be created ?
[EDITED] – Advanced out-of-the-box deployment without registering a voice profile is also available.
The purpose of the personal voice profile
The personal voice profile contain acoustic information including calculated and statistic data on many parameters that characterize the voice of the registered speaker. In a noisy environment, the personalized noise reduction is using the information that is available in the profile in order to differentiate between the registered speaker and the noise so it can provide a clean audio that contains the registered voice with attenuated noise.
What is required in order to build the personal voice profile ?
In order to build a personal profile, the algorithm should get as input few minutes of the speaker’s voice that were captured in a quiet environment. There is no need for tedious training like the need to repeat specific sentences. This technology is flexible and can use free-style voice recording of the speaker.
The registration process
On each system, the registration process, in which you create the voice profile from the audio, should be configured to be as transparent as possible for the end users. One option is to record few minutes of the voice during normal calls that are made from a quiet environment. This process is done during the call itself, by streaming the incoming audio to SoliCall.
Alternatively, many organization, like call centers, banks etc., already store recorded audio of their employees/agents. In this case the registration process can be done offline by filtering the recorded files.
No matter which technique is used, the registration process can be done in a way that is transparent to the end users.