Correcting some misconceptions about Voice Biometrics for Forensics - Sestek - ContactCenterWorld.com Blog
As I have explained in the previous post, Voice Biometrics is more reliable than fingerprint in some aspects. Analyzing voice based evidence is becoming a popular tool for law enforcement and criminal identification. However, there are some misconceptions about terminology and also the limits of what is currently possible with Voice Biometrics technologies. Let me explain some of the terminology to prevent misconceptions.
Likelihood Ratio (LR): LR is calculated by comparing two audio files one-to-one. Based on this comparison, we generate statistical scores and probability distribution graphics as evidences of similarity or differences.
Speaker Identification: Speaker Identification constitutes an effective solution for forensic voice analysis by confirming/denying the identity of individuals based on audio files that are used during investigations.
Voice Biometrics technologies cannot identify the voice of an individual from an unlimited set of voice samples. The model voice is usually compared against a pre-defined set of voices that are already identified and/or labeled.
This set of voices (reference population) is used for normalization of the scores. 20 is the minimum size of the population sample for normalization. Increasing the size beyond 50 does not improve the normalization performance. Therefore, the optimum size for populations is between 20 and 50.
To have reliable scores, the model voice and pre-defined set of voices should match in terms of gender, language and input source characteristics. For example, it does not make sense to compare a female voice with a set of male voices.
Let me know if you have questions. I will be glad to answer them and/or think along with you.
Publish Date: November 9, 2016 5:00 AM
|All Suppliers||Get Listed|
(VIEW OUR PAGE)
Advanced AI technology and Natural Language Processing delivered to clients in the Cloud that harnesses both voice and digital conversations. The focus is on building an environment where intelligent ...
(VIEW OUR PAGE)
Аутсорсинговый контакт-центр ConceptCall| КонцептКолл специализируется на исходящем и входящем телемаркетинге: мы предоставляем услуги по осуществлению холодных звонков и продажи по телефону, проводим...
(VIEW OUR PAGE)
CTI Software is the creator of custom application called Davos, which complements the telephone client solutions with intelligent features such as automated call attendant, call recording and archivin...
View more from Sestek
Recent Blog Posts:
|Using Voice Biometrics in Law Enforcement: Audio Forensics||January 20, 2019 5:00 AM|
|Engage with Your Customers Smartly: Use the Power of Conversational Technologies||December 4, 2018 5:00 AM|
|Fighting Against Call Center Fraud with Voice Biometrics||November 13, 2018 5:00 AM|
|Enjoying Self-Service Benefits with Speech Enabled IVR||August 15, 2018 5:00 AM|
|Four Steps to Providing the Virtual Assistant Your Customers Ask For||September 5, 2017 5:00 AM|
|Introduction to Chatbots||May 2, 2017 5:00 AM|
|Top Technology Trends in 2017 and their Intersection with Speech Technologies||January 19, 2017 5:00 AM|
|Yin-Yang in Customer Interactions||November 28, 2016 5:00 AM|
|Correcting some misconceptions about Voice Biometrics for Forensics||November 9, 2016 5:00 AM|
|Speech Analytics at SESTEK||November 9, 2016 5:00 AM|