結合平移距離與傅立葉特徵之手勢辨識系統

本篇論文提出一套以數位影像為基礎的手勢辨識系統，用於辨識出七種已知的手勢，系統的架構分成三大模組，第一個模組是前處理模組、第二個模組為特徵擷取模組、最後一個模組則是辨識模組。系統在發展的過程當中，以距離特徵向量為基礎的辨識器與以FFT特徵向量為基礎的辨識器所得到的辨識結果不甚理想，最後使用一種組合演算法以權重的方式結合兩者，此演算法結合了兩者個別的優點亦去除兩者的個別的缺點，並也得到了較好的成果。在系統的實驗成果當中，如果測試影像是已經訓練過的資料其所得到的辨識正確率為95%，而如果測試影像不在訓練資料當中，所得到的正確率為87%，因此本套手勢辨識系統可以適用在未來用於加強人與機器之間的互動式系統。

關鍵字

圖形識別；手勢；膚色偵測；影像處理

並列摘要

A hand-pose recognition system is presented. Our system is image-based and designed to recognize seven different hand-poses in three major phases, i.e., pre-processing, feature extraction, and recognition. At the early recognition, two “weak” classifiers that are based on either shift-distances or fast Fourier transform (FFT) features are developed. The results are further combined to form a “strong” classifier. Our experimental results indicate that the system has achieved preliminary detection rates of 95% for the training set and 87% during testing. In summary, we have presented a system that could be used to enhance the human-computer interaction in daily applications.

並列關鍵字

skin color detection ； Hand-pose ； pattern recognition ； image processing

參考文獻

Google Scholar

[1] H. Sheikhzadeh and L. Deng, “Waveform-Based Speech Recognition Using Hidden Filter Models: Parameter Selection and Sensitivity to Power Normalization,” IEEE Trans. on Speech and Audio Processing, Vol. 2, no. 1, pp. 80-89, Jan. 1994.

Google Scholar

[2] P. Woodland, “Speech recognition”, Speech and Language Engineering - State of the Art, IEEE Colloquium on, 19 Nov. 1998.

Google Scholar

[3] C. Y. Suen, K. Kim, Q. Xu, J. Kim, and L. Lam, “Handwriting Recognition – The Last Frontiers,” icpr, p. 4001, 15th International Conference on Pattern Recognition (ICPR'00), Vol. 4, Sept. 2000.

Google Scholar

[4] S. N. Srihari and S. Lee, “Automatic handwriting recognition and writer matching on anthrax-related handwritten mail,” Frontiers in Handwriting Recognition Proceedings. Eighth International Workshop on, pp. 280-184, 2002.