Deng Jiang, Vice President of Zhongguancun Science and Technology: Practical application of AI audio and video technology in financial scenarios

The traditional financial industry faces pain points such as inefficient operations, poor risk management and control, and high customer acquisition costs. Solving traditional business pain points and difficulties through the integration of finance and technology is an important support for the current innovative development of the financial industry.

Recently, at the AISummit Global Artificial Intelligence Technology Conference hosted by 51CTO, Deng Jiang, Vice President of Zhongguancun Science and Technology, gave a keynote speech "Practical Application of AI Audio and Video Technology in Financial Scenarios", introducing the technical application and thinking of audio and video technology and finance from three levels: application, technology, role and value. .

In the past three years, the epidemic has had a great impact on the entire economy and society. Financial businesses that rely heavily on offline scenarios are no exception.

Application of AI audio and video technology in financial scenarios

In the context of the epidemic in the past few years, the financial business has been greatly affected As a result, the country has also introduced a series of management measures to promote the development of contactless finance.

Under the requirements of the new environment and driven by new technologies, the traditional human-driven service model (offline human-driven service model) is iterated into AI-driven intelligent services Model (service model of online and offline omni-channel human-machine system). Under the traditional human-driven service model, only offline face-to-face, text, telephone voice, audio and video and other service modes can be realized. Under the AI ​​driver, it can realize intelligent/unmanned outlets, intelligent customer service, intelligent IVR/outbound calls, and AI intelligence. Video, AI virtual employee and other services.

In order to realize remote banking, Deng Jiang said that there are five core technologies that are driving the progress of the entire technology. The five core technologies include artificial intelligence technology, real-time computing, biometric identification and identity verification, data decision-making and data computing, and privacy protection.

In the implementation process, there are three core algorithms and four core technical capabilities, which are the other two parts of technical needs in the process of technology implementation. Three major AI core algorithms: speech technology, natural language processing, and machine vision. Four core technical capabilities: omni-channel high-quality audio and video communication capabilities, omni-channel SDK packaging and adaptation capabilities, audio and video-based AI algorithm deep integration and application capabilities, and flexible and visual video service scene orchestration capabilities.

Technical practical operation of intelligent video cloud construction in financial scenarios

Deng Jiang said that intelligent video cloud is a way to promote basic video Digital upgrade, using AI intelligence and RPA process automation, to build a new video service model of "human-machine collaboration and human-machine self-service". With the support of basic cloud computing resources, the bottom layer builds an audio and video platform that supports high concurrency and fast response, including ASR, TTS, NLP, OCR, face recognition, anti-remake, and live detection, etc. At the business middle-end level, it implements customer process management, intelligent queuing under high concurrency, statistical analysis of relevant information, order management, and a series of middle-end support.

# On the front end, in addition to the support of multi-end intervention, a series of multi-modal biometric verification for counterfeit identities , client self-service and customer service remote video, as well as real-time calculation and capture of on-site video. The front end is the business scenario end, which includes related online and offline integrated process management for a series of businesses such as financial management, insurance, trust, etc.

Relying on a strong underlying foundation, five core products have been formed: multi-modal identity verification, AI on-site video service, AI self-service video service, AI remote video service, and AI intelligent audio and video quality Check. It also introduced the video service system of AI Video Cloud, dismantled the product system, and introduced product features, process management, and product value.

Practical application of AI audio and video technology in financial scenarios - mobile credit

There are six major difficulties in traditional financial credit interviews:

Faced with information islands, interview data is independent of the risk control system and does not fully utilize its dynamic data value; it is a fully manual-driven model, and the quality is highly dependent on personnel experience and is uneven; business efficiency is low, and there is no intelligence or weak intelligence. Auxiliary, high pressure at the grassroots level, low efficiency; single business exhibition channel, on-site interview and interview mode, difficult to cover and high cost; business volume bottleneck, business peaks and troughs, poor dynamic expansion matching; manual sampling risk is high, manual offline sampling review, potential The risks are high, feedback is not timely enough, and personnel work under high pressure.

After Deng Jiang introduced the difficulties in interviewing, he also interpreted four industry "Notices" issued by the China Banking and Insurance Regulatory Commission from a policy level. Deng Jiang said that leaving audio and video traces has become a mandatory requirement in the banking, trust, insurance, and securities industries.

In Mr. Deng Jiang’s speech, he shared four scenarios of mobile credit: remote video interview, self-service video Interviews, door-to-door interviews with account managers, and on-site interviews at outlet counters were also introduced. The entire video risk control process and the practical results of Zhongguancun Science and Technology in the direction of bio-anti-counterfeiting, , are multi-modal bio-anti-counterfeiting and security platforms.

The multi-modal biological anti-counterfeiting and security platform supports multiple living detection methods such as motion and reading, and uses the powerful AI algorithm on the server to provide more accurate identification and response. For fraud capabilities, its platform is divided into four layers, consisting of access layer, core layer, functional layer, and scenario layer.

The access layer consists of WeChat applet, APP, mobile H5, Web, camera port and third-party system .

The core layer has three functional modules, including live counterfeiting, fraud detection and biological comparison. The live counterfeiting function is composed of basic and enhanced anti-counterfeiting detection and behavioral risk detection. Basic anti-counterfeiting detection includes face presentation attack, voiceprint presentation attack, and link hijacking detection; enhanced anti-counterfeiting detection includes voice migration synthesis detection, Deep counterfeit detection; behavioral risk monitoring includes face posture detection, lip language recognition, audio and video synchronization detection, and occlusion semantic segmentation. Fraud detection includes ID card forgery detection, signature and seal forgery detection, portrait background similarity, and voiceprint gang discovery. Biological comparison includes adversarial sample enhanced learning, voiceprint comparison retrieval, and face comparison retrieval.

The functional layer consists of verification capability assessment, behavioral risk assessment, policy management, third-party data access, federated learning, active attack interception, encrypted storage, senseless registration, senseless recording, 12 modules including sensorless refresh, life cycle management, and security audit are used to implement functions.

The business scenarios included in the scenario layer include multi-dimensional real-name authentication, credit granting, employee compliance supervision, intermediary agent detection, detection of electronic review gangs, CC complaint tracing, office desktop security, Big customers come online.

After an in-depth analysis of the multi-modal biological anti-counterfeiting platform, the platform function of AI intelligent audio and video quality inspection and manual spot inspection and review was explained. With the help of AI visual and voice quality inspection technology, In the video service, real-time quality inspection, real-time correction (text correction, voice correction), real-time reminder to users and business managers, greatly improving the first pass rate and avoiding the problem of high cost of secondary registration and poor experience for users, the main technologies involved include Intelligent image recognition, intelligent biometric identification, intelligent voice recognition, intelligent action recognition, audio and video synchronization detection, etc.

In the introduction of intelligent collection and intelligent return visits, intelligent collection can realize fully automated collection operations and anthropomorphic communication; the speech can be flexibly customized for different overdue stages and customer types; the standard speech process can avoid manual labor Compliance risks and complaints caused by irregular collection techniques. Intelligent return visits can achieve high call efficiency; understand customer reach rates through background statistics; be enthusiastic and full without affecting customer experience; reduce costs and increase efficiency.

After analyzing the functions and technologies involved in the AI ​​intelligent video cloud, we shared relevant cases of the AI ​​intelligent video cloud. Please see the video playback on the official website for details of the case.

Conclusion

In AI audio and video technology, whether it is face, voiceprint, lip reading, speech synthesis, etc. There is in-depth scene customization in the scene. As a technology company, in addition to polishing technical capabilities, it is more about in-depth business scenarios, being customer-centric, understanding customer needs, solving real pain points in customer business, and being able to make good use of tools. This is the future of technology. The company's higher requirements. Ultimately, through the in-depth application of technology in financial scenarios, the level of the entire financial business will be improved and the boundaries of large-scale development of the entire financial business will be broadened.

The conference speech replay and PPT are now online. Go to the official website to view the exciting content--> ​​AISummit Global Artificial Intelligence Technology Conference Official Website​​

The above is the detailed content of Deng Jiang, Vice President of Zhongguancun Science and Technology: Practical application of AI audio and video technology in financial scenarios. For more information, please follow other related articles on the PHP Chinese website!