Speaker Verification System

Speaker Verification Submenu

- high level description
- preprocessing
- feature extraction
- enrollment
- threshold creation
- verification


Signal preprocessing is a packaging for to function calls. The first is silence removal, the second is a high pass filter.

Silence Removal
If there was silence removal, part of the voiceprint would come from the silence and an imposter could easily pass by just recording silence. So, it is imperative that the silent parts at the beginning and the end of the recording be deleted. The first 0.05 seconds of the recording is treated as silence. A silence threshold is generated by finding the maximum amplitude in the first 0.05s. The program then finds the first time in the recording where the amplitude exceeds the threshold and uses that as the starting index to the recording. Silence at the end of the file is removed using the same threshold and is identified by the last time the recording drops below the threshold.

High Pass Filter
A high pass filter was used to filter out the low frequency background noise. The cutoff frequency of the filter is ~1900Hz. A frequency response plot is in figure 2 below.

Figure 2: Preprocessing Filter Frequency Response