Biometric - derived from the Greek words "Bio" (life) and "Metric" (measurable), refers to the calculation, measurement, and comparison of unique physiological or behavioral characteristics of individuals to identify their identity. In other words, your DNA and personal behaviors (habits) are the password.

Static - Physiological Characteristics	Fingerprint, facial features, iris, veins, palm print, voiceprint, etc.
Dynamic - Behavioral Characteristics	Voice, heartbeat, gait, signature, keyboard typing speed & pressure, etc.

Voice satisfies both physiological and behavioral characteristic requirements:

Physiological Characteristics: Factors influencing audio include throat, mouth, nose size, vocal cords, and the length of the vocal tract.
Behavioral Characteristics: Each individual has different habits regarding speaking volume, speed, intonation, etc.

What information can be obtained from a voice call?

What is the content of the call? (NLP technology)
What language is spoken during the call?
Who is speaking during the call? (Voice Biometric)
The gender of the speaker during the call.
The age of the speaker during the call.
The emotions expressed by the speaker during the call.

Voiceprint Scene Classification Based on Text:

Text-Dependent: Completely restricted to specified voice content, for example, asking customers to read "I love NanXun VoiceID" or "My voice is my password."
Text-Related: The voice content must fall within specific ranges. For example, it could be limited to numbers or colors.
Text-Independent: The voice content can be anything.

NSE Verint Voiceprint Technology Application Scenarios - Verification vs. Identification:

Customer Identity Verification (shorten authentication time/improve security/optimize interactions).
Fraudster Identification (enhance risk compliance/reduce fraud losses/upgrade security strategies and customer interaction processes).

Identity Verification (1:1)	Identity Identification (1:N)
"Are you really who you claim to be?" Return: Yes or No	"Who are you?" Return: Not found or [the identity system deems as the most matching identity]

Biometric Recognition Accuracy - False Rejection Rate vs. False Acceptance Rate:

Type I Error - FRR (False Rejection Rate):

(System's errors rejecting legitimate users' attempts to access the system ÷ legitimate users' attempts to access the system) x 100%

FRR = 0.1% means for every 1000 legitimate user attempts to log in, the system will falsely reject 1 legitimate user.

A higher FRR value indicates lower system convenience, so we aim for a lower FRR.

Type I Error - FAR (False Acceptance Rate)

(System's errors accepting fraudsters' attempts to access others' systems ÷ fraudsters' attempts to access others' systems) x 100%

FAR = 0.1% means for every 1000 fraudsters' attempts to access others' systems, the system will falsely accept 1 fraudster.

A higher FAR value indicates lower system security, so we aim for a lower FAR.

NSE Verint Voiceprint Recognition Advantages:

Real-time voiceprint recognition during the recording process on the recording server.
Encryption and tamper resistance to prevent substitution attacks on voiceprints.
Ability to perform voiceprint recognition using historical recordings (Archive Campaign).
Capability to perform voiceprint registration using historical recordings.
Ability to import external audio files for voiceprint registration.
Voiceprint recognition data is input into the Data Warehouse for subsequent reporting.
Evolution of underlying algorithms based on actual results to enhance predictive analysis models (high detection rate/low false positive rate).