查詢結果分析
來源資料
頁籤選單縮合
| 題 名 | 人工智慧時代的筆跡鑑定:基於AI反向學習與人機合作策略=Handwriting Examination in the Age of Artificial Intelligence: AI-Based Reverse Learning and Human-Machine Collaboration Strategies |
|---|---|
| 作 者 | 高信雄; | 書刊名 | 警學叢刊 |
| 卷 期 | 55:4=275 2024.12[民113.12] |
| 頁 次 | 頁1-20 |
| 分類號 | 312.831 |
| 關鍵詞 | 簽名驗證; 深度學習; 人機協作; 可解釋性AI; AI反向學習; Signature examination; Deep learning; Human-machine collaboration; Explainable AI; XAI; AI reverse learning; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 本研究探討了可解釋性人工智慧技術 (XAI) 在筆跡鑑定上的應用與發展潛力,並 初步嘗試探討AI 反向學習的可能性。本研究基於深度卷積神經網路 (DCNN) 和獨特 的局部特徵擷取方法,提出了一種離線手寫簽名驗證方式,並以單一已知簽名為樣本 進行驗證,在兼顧準確率的同時,也確保此方法具備較高的實用性。同時我們採用了 具有可解釋性的模型,進一步以圖像輔助鑑識人員判斷簽名真偽。在對實驗的分析部 分,本文探討了人類對AI 反向學習的可能性。我們發現這種學習方式有一定的優勢, 有利於打破人類僵化思維模式,更可藉由AI 進行量化分析以提供更可靠的鑑定結果。 最後,本研究提出了一種由淺入深的「人機協作階梯模型」,涵蓋了以AI 為輔助工 具、學習AI 行為、理解AI 決策過程,最終達到人類專家和AI 的人機智慧融合,促 成更高效的協作。本文認為,善用此種AI 協作模式,將有效提升警政工作的效能,亦 是警察人員在人工智慧時代習知識或掌握新技能的契機。 |
| 英文摘要 | This study investigates the application and developmental potential of humaninterpretable artificial intelligence (AI) technologies in handwriting examination. It also explores the possibilities of AI reverse learning. Leveraging deep convolutional neural networks (DCNNs) and a unique method for local feature extraction, we propose an offline handwritten signature examination approach that validates against a single known signature sample. This method not only maintains accuracy but also ensures high practicality. Furthermore, we utilize Explainable AI (XAI) models to aid forensic experts in judging the authenticity of signatures, enhanced by visual aids. Through analyzing our experiments, this paper discusses the potential for human reverse learning from AI. We find that this learning modality has distinct advantages, aiding in disrupting rigid human cognitive patterns and providing more reliable forensic results through AI's quantitative analysis. Finally, we introduce a 'Human-Machine Collaboration Ladder Model' that progresses incrementally, covering the use of AI as an assistive tool, learning AI behaviors, understanding AI decision-making processes, and ultimately achieving a synergistic integration of human expertise and AI. This paper posits that effective utilization of this AI collaboration model will significantly enhance the efficiency of police administration work and presents an opportunity for law enforcement personnel to acquire knowledge or master new skills in the age of artificial intelligence. |
本系統中英文摘要資訊取自各篇刊載內容。