A business flow diagram of Customer Identification via Video Call
Explanation
04 steps to identify customers via Video Call
1. Customers use the App to fill in the information
- Customer Information is automatically received using OCR to extract information from identification.
- Customers use the App to directly capture their identity papers, limiting the ability of customers to upload edited photos.
- Customers have the option to manually enter personal information
2. The customer confirms the information in the Account/Service registration form and sends it to the eKYC system
Customers check filled information is correct and sends it to the eKYC system
3. (Optional) The customer sets an appointment to perform identification via Video Call
- The customer's identification schedule will be sent to the eKYC system and notified to the agent
- Customers can change the schedule, reschedule if they cannot make Video Call at the right time
4. Make a Video Call or receive a call to conduct an identification
- In case the customer actively calls:
- Calls will be queued and distributed to available agents.
- In case the customer receives the call:
- The call will be ringed actively in the Mobile App that the customer has logged in.
- Customers do not need to open the App continuously.
Components of eKYC
- App for Customers:
- Mobile App: iOS/Android App with Video Call SDK integrated (Mobile App is recommended)
- Web App: Customers using Laptop/PC can access the Web site that has integrated Video Call SDK
- Constructed by the management unit under the advice of the Vendor
- Video Call SDK provided by the vendor
- App for Agent:
- Web App: Web users have integrated Video Call SDK and connected to AI system, eKYC system by API
- Construction is carried out by the management unit under the advice of the Vendor, or the Vendor builds and customizes according to business requirements.
- Video Call SDK provided by Vendor
- AI system:
- Be responsible for handling the following tasks:
- Extract information from identity papers (OCR)
- Compare the match of portrait photos on paper and actual portraits (Face-matching)
- Be built and deployed by vendor; communicate with other components in the business flow using RESTful API
- Be responsible for handling the following tasks:
- eKYC system:
- As a customer identification system as a management unit, built and managed
- Connect to AI system and application using API
- Store customer information
- Log, record, a record of calls
Specifications:
A. SDK Compatibility
- Mobile App:
- Android: Version 4.1+
- iOS: Version 10.0+
- React Native: According to the platform's native operating system
- Web App:
- SDK Javascript
- Requires a browser that supports WebRTC
- Recommends using Chrome
B. Internet bandwidth
Request by Internet bandwidth 01Mbps for 01 stream Video Request by Internet bandwidth 01Mbps for 01 Video stream
- The case requires the Customer and Agent to see each other's Video: 02 Mbps for each party.
- The case only requires Video from the Client side: 01 Mbps for each party.
Resolution: 720p 30 FPS
SDK Video Call supports Adaptive Quality: Video quality can be reduced to the resolution of 360p in cases of low or unstable Internet bandwidth. However, the minimum required bandwidth is 0.5Mbps.
C. Storage
- Log files, audio recordings, and video recordings are stored at:
- Vendor's Video Call Infrastructure
- eKYC system
- Video File Size: 10-12 MB/min (Bitrate 3Mbps)
Feature Commitment
Video Call System
Features | Availability |
Real-time 2D video recording | OK |
Real-time 2D audio recording | OK |
Capture images in Video | OK |
Call encryption | OK |
On-Premise Deployment | OK |
High Availability | OK |
Load Balancing | OK |
AI System (OCR and Face Matching)
Features | Provision Availability |
Ability to compare, contrast and provide matching image results on the customer's identity papers and selfies | OK |
Supported identity papers: - ID Card (Both old and new types) - Passport | OK |
Extract information in Vietnamese with stamps from identity papers | OK |
Detect/ warn users of un-uploaded or blurred images | OK |
Invalid status document detection/ warning includes: - Fake - Edited - Corner cut, hole punched - Blurred - Expired | OK (need to integrate the national database, or the bank itself to compare and find forged or altered documents) |
Support to identify fake/duplicate identity papers by comparing taken and uploaded customer image data with available data inside and outside the business (customer's image coincides with another ID card or with another individual's information in external data) | OK |
Identify basic image file formats JPG, PNG, GIF | OK |
Identify and warn users when the uploaded image and the registered image are different | OK |
Identify customers as existing customers through images and simple actions | OK |
Can be Integrated with other applications, platforms | OK |
Deploy on App/Web platform | OK |
Load balancing | OK |
Meet security requirements | OK |
Face-matching accuracy between identity papers and selfie | OK |
False acceptance rate (FAR) for each type of identity paper | 0.9% |
False rejection rate (FRR) for each type of identity paper | 3.5% |
SDK: - Allows identification of real people or ready-made photos. - Support Android, iOS - Full support for popular models. | OK |
The system automatically recognizes the type of identity papers without manual selection | OK |
Extract the information fields of the ID card on the pdf file | OK |