Auditory brainstem reaction automatic classification method, device and equipment based on time-frequency cross attention network and storage medium

By employing a time-frequency cross-attention network-based approach, ABR signals are processed in a dual-branch parallel manner, overcoming the limitations of feature representation and multi-domain fusion in existing technologies. This approach enables high-precision automatic ABR classification, improving the accuracy and efficiency of hearing screening and diagnosis.

CN121901936AActive Publication Date: 2026-04-21THE CHINESE UNIV OF HONG KONG (SHENZHEN)
3 Cites 0 Cited by

Patent Information

Application Number
CN202610362315.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-24
Publication Date
2026-04-21
Estimated Expiration
2046-03-24

AI Technical Summary

Technical Problem

Existing ABR automated analysis technology has limitations in feature representation, utilization of prior information, and multi-domain fusion mechanisms, making it difficult to meet the clinical needs for high accuracy, robustness, and physiological interpretability. In particular, the interpretation results differ under low signal-to-noise ratio or atypical waveform conditions, and traditional manual interpretation is time-consuming and difficult to standardize.

Method used

A method based on time-frequency cross-attention network is adopted to perform dual-branch parallel processing on ABR signal. The time-domain branch extracts local waveform morphology features through one-dimensional convolution, and the frequency-domain branch obtains time-spectrum features through short-time Fourier transform. Feature linear modulation mechanism is introduced and position embedding is added. Deep fusion of time-frequency features is achieved through bidirectional cross-attention module. Finally, global average pooling and multilayer perceptron are combined for classification decision.

Benefits of technology

It achieves high-precision automatic classification on clinical ABR data, significantly improving the accuracy, objectivity and efficiency of hearing screening and diagnosis. It eliminates the need for manual feature engineering and can make full use of multimodal information and physiological priors.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses an auditory brainstem response automatic classification method and device based on a time-frequency cross attention network, equipment and a storage medium, and relates to the technical field of hearing monitoring. Double-branch parallel processing is carried out on ABR signals; the frequency domain branch obtains time-frequency spectrogram features through short-time Fourier transform; a characteristic linear modulation mechanism is introduced into each of the two branches, and stimulation intensity is used as conditional information to carry out adaptive scaling and offset on intermediate characteristics; complementary fusion of the time-frequency characteristics is realized through a bidirectional cross attention mechanism, so that the time-domain characteristics can query the frequency-domain information, and the frequency-domain characteristics can query the time-domain information; and performing binary classification decision by combining global average pooling, stimulation intensity late fusion and a multi-layer perceptron. According to the method, multi-modal information and physiological priori can be fully utilized in automatic classification without manual feature engineering, and the accuracy, objectivity and efficiency of hearing screening and diagnosis are remarkably improved.
Need to check novelty before this filing date? Find Prior Art