Ping An Technology Applies for "AI-Based Emotion Recognition" Patent to Enhance Fintech Efficiency and Accuracy

Ping An Technology Applies for "AI-Based Emotion Recognition" Patent to Enhance Fintech Efficiency and AccuracyOn December 5, 2024, the State Intellectual Property Office of China revealed that Ping An Technology (Shenzhen) Co., Ltd

Ping An Technology Applies for "AI-Based Emotion Recognition" Patent to Enhance Fintech Efficiency and Accuracy

On December 5, 2024, the State Intellectual Property Office of China revealed that Ping An Technology (Shenzhen) Co., Ltd. submitted a patent application titled "Method, Apparatus, Computer Equipment and Medium for AI-Based Emotion Recognition," with publication number CN119068914A and an application date of August 2024. This patent application focuses on the intersection of artificial intelligence and fintech, aiming to improve the efficiency and accuracy of speech emotion recognition. The patent abstract details the core methods and application scenarios of this technology.

The core of this patent application lies in an AI-based emotion recognition method. This method cleverly combines teacher and student models. Through training and optimization, it achieves accurate emotion recognition of target voice data. Specifically, this method involves the following key steps:

First, the system calls both the teacher and student models. The teacher model, acting as an experienced "mentor," guides the learning and training process of the student model. The student model, as the learner, continuously absorbs knowledge and experience from the teacher model, ultimately achieving performance equal to or exceeding that of the teacher model.

Second, the system uses acquired voice sample data to train the teacher model. During this training, the system obtains joint embeddings and training outputs, generating the teacher model's first loss. The first loss represents the error produced by the teacher model during training, reflecting its performance level.

Simultaneously, the system uses the same voice sample data to train the student model. The goal of this training is to obtain specified voice embeddingstransforming voice signals into feature vectors understandable by the model.

Next, the system generates the student model's second loss based on the joint embeddings and voice embeddings. The second loss reflects the error produced by the student model during training, indicating its performance level.

To comprehensively consider the performance of both models, the system generates a combined loss based on the first and second losses. This combined loss represents the overall error of the entire training process, which the system uses to optimize the student model.

After multiple iterative optimization processes, the system ultimately obtains a target emotion recognition model. This model possesses high-efficiency and high-accuracy emotion recognition capabilities, accurately identifying emotional information in target voice data.

Beyond the emotion recognition method, the patent application also covers an AI-based emotion recognition apparatus, computer equipment, and storage media. These provide the hardware and software platform for practical application of the emotion recognition method.

It's noteworthy that the patent application proposes storing the emotion recognition results on a blockchain. Blockchain technology ensures the immutability and traceability of the results, further enhancing system security and reliability.

In summary, Ping An Technology's patent application for the "Method, Apparatus, Computer Equipment and Medium for AI-Based Emotion Recognition," with its innovative technical solutions and broad application prospects, is expected to play a significant role in the fintech field. By combining teacher and student model training methods and utilizing blockchain technology, this patent application effectively improves the efficiency and accuracy of speech emotion recognition processing. This is critically important for financial institutions to better understand customer emotions, improve service quality, and manage risks.

This technology can be applied to various financial scenarios, such as: voice interaction analysis in customer service centers (adjusting service strategies based on customer emotion); risk assessment and management (identifying potential risks through analysis of customer voice emotion during transactions); and anti-fraud monitoring (identifying potential fraud through abnormal emotional fluctuations in voice).

The success of this patent application demonstrates Ping An Technology's continuous innovation in AI and fintech, foreshadowing a future where fintech is more intelligent and human-centric. The mature application of this technology will bring more accurate and convenient service experiences to the financial industry, while also providing more reliable security.

The innovation lies in the clever combination of teacher and student models and the optimization of the student model through a combined loss function, effectively improving the accuracy and efficiency of emotion recognition. Furthermore, storing emotion recognition results on a blockchain enhances data security and reliability, providing a solid foundation for applications in the fintech field. This is not only a technological breakthrough but also a significant milestone in the development of the fintech industry. Ping An Technology's patent application will undoubtedly drive the financial industry towards a more intelligent and secure future. We can expect to see this technology applied in more financial scenarios, bringing more efficient, convenient, and secure financial services to customers.