Horizon

Voice Recognition System (VRS)

Product Overview

Monitoring incoming & outgoing calls and detecting threats/fraudulent activities is the essential need of National Security. Moreover, to verify the identity of a speaker during the course of natural conversation is important. Horizon Tech Services provides biometric speaker recognition technology using the voice features unique to a particular speaker for authentication solution. Totally content, language and accent independent, the system transparently retrieves the biometric voice characteristics required for verification within seconds. Moreover, it enables to generate/send alarm to respective client.

Key Features

Pre-Processing

  • Identify and separate DTMF, silence, noise and speech
  • Automatic noise suppressor and frequency compensator
  • Speaker segmentation

Training

  • Extracts MFCC and Pitch features of target speaker and create Voice Print stored in database
  • Recommended minimum SNR 12-15 dB
  • Recommended minimum duration 60 sec

Identification

  • Automatically compares and identifies speech sample against suspects’ speech clips
  • Use probabilistic method to give the matched result in likelihood ratio
  • Millions of voiceprints (VP) compared within a min

Verification

  • Automatically 1-1 verification of speech sample based on 16 distinct pitch parameters
  • The matched speech samples results are normalized and compared with the preset threshold to accept or reject a speaker.

Report Generation

  • Generates the report based on the acquired matched Identification/ verification results.
  • Log files are maintained for all search tasks
  • Generates Call related information .e. calling ID, dialed number, time and duration and Link chart analysis report

System Configuration

  • Standalone and Server-Client
  • Process real time or recorded audio data for identification/verification of speaker