正常人左右腳新特徵探索 Round 1

使用 ASR10 + FIR0.5-50 + ICA 後切割正常人資料,避開前面已做過的 bandpower、MRCP、Hjorth、permutation entropy、channel correlation、CSP 等特徵,先嘗試 15 種不同角度的 left/right 特徵。

讀法

每張圖左側為 32 通道 left-right paired Cohen dz 腦拓圖,右側為該特徵最佳通道的正常人 left/right paired plot。這一輪是探索性篩選,最佳通道尚未固定且未做多重比較校正;目標是先找出值得下一輪驗證的候選特徵。

特徵排名

Feature ranking
特徵最佳通道|dz|p方向一致性判斷
Mu ITPC 8-13 HzC41.3860.00180.90promising
Lag-1 autocorrelationP71.0770.00780.90promising
Late - early line lengthT80.9120.01800.80promising
Aperiodic slope 5-40 HzF70.8520.02470.80watch
Zero-crossing rateF70.8330.02720.80watch
Lempel-Ziv complexityF70.7960.03300.80watch
Absolute peak latencyP30.7830.03520.80watch
Spectral edge 90% 5-30 HzF70.7550.04060.80watch
Spectral centroid 5-30 HzF70.7240.04780.70watch
Low-beta ITPC 13-20 HzC30.7180.04940.70watch
Alpha peak frequencyP30.7100.05150.80watch
Higuchi fractal dimensionF70.7060.05260.80watch
Spectral entropy 5-30 HzF70.6950.05540.70watch
Teager-Kaiser energyFC40.6330.07630.80weak
Line lengthFC40.6120.08490.80weak

Line length weak

不是 MRCP amplitude/slope/AUC,也不是 Hjorth 或 permutation entropy;它看的是波形總變動量。

Best channel FC4,|dz|=0.612,p=0.0849,consistency=0.80

Line length feature plot

feature note

Zero-crossing rate watch

不是振幅或 PSD,主要看 0~2 秒內訊號穿越均值的頻率。

Best channel F7,|dz|=0.833,p=0.0272,consistency=0.80

Zero-crossing rate feature plot

feature note

Teager-Kaiser energy weak

和 bandpower 不同,TKE 對瞬時振幅與頻率共同變化敏感。

Best channel FC4,|dz|=0.633,p=0.0763,consistency=0.80

Teager-Kaiser energy feature plot

feature note

Lag-1 autocorrelation promising

不是 sliding channel correlation;這裡看單一通道自身的一階時間記憶。

Best channel P7,|dz|=1.077,p=0.0078,consistency=0.90

Lag-1 autocorrelation feature plot

feature note

Late - early line length promising

加入 0~1 秒與 1~2 秒的時間窗差,測試運動期間後段是否有額外變動。

Best channel T8,|dz|=0.912,p=0.0180,consistency=0.80

Late - early line length feature plot

feature note

Absolute peak latency watch

不是 MRCP 振幅/斜率/面積;它只看最大絕對偏移出現的時間。

Best channel P3,|dz|=0.783,p=0.0352,consistency=0.80

Absolute peak latency feature plot

feature note

Lempel-Ziv complexity watch

不同於 permutation entropy;LZ 看二值化序列中新增片段的壓縮複雜度。

Best channel F7,|dz|=0.796,p=0.0330,consistency=0.80

Lempel-Ziv complexity feature plot

feature note

Higuchi fractal dimension watch

不同於 Hjorth mobility/complexity;Higuchi FD 估計時間序列幾何粗糙度。

Best channel F7,|dz|=0.706,p=0.0526,consistency=0.80

Higuchi fractal dimension feature plot

feature note

Spectral centroid 5-30 Hz watch

不是頻帶 power 大小,而是 5~30 Hz 頻譜的重心位置。

Best channel F7,|dz|=0.724,p=0.0478,consistency=0.70

Spectral centroid 5-30 Hz feature plot

feature note

Spectral entropy 5-30 Hz watch

不是 permutation entropy;它衡量頻譜能量分布是否集中或分散。

Best channel F7,|dz|=0.695,p=0.0554,consistency=0.70

Spectral entropy 5-30 Hz feature plot

feature note

Spectral edge 90% 5-30 Hz watch

不是 bandpower,而是 90% 累積頻譜能量所在頻率。

Best channel F7,|dz|=0.755,p=0.0406,consistency=0.80

Spectral edge 90% 5-30 Hz feature plot

feature note

Aperiodic slope 5-40 Hz watch

不是固定頻帶功率;它估計 log PSD 對 log frequency 的斜率。

Best channel F7,|dz|=0.852,p=0.0247,consistency=0.80

Aperiodic slope 5-40 Hz feature plot

feature note

Alpha peak frequency watch

不是 alpha power,而是在 8~13 Hz 內找峰值頻率位置。

Best channel P3,|dz|=0.710,p=0.0515,consistency=0.80

Alpha peak frequency feature plot

feature note

Mu ITPC 8-13 Hz promising

不是 power 或 channel correlation;它看 trial 間相位是否鎖定。

Best channel C4,|dz|=1.386,p=0.0018,consistency=0.90

Mu ITPC 8-13 Hz feature plot

feature note

Low-beta ITPC 13-20 Hz watch

不是 beta power;它測試 low beta 相位鎖定是否左右腳不同。

Best channel C3,|dz|=0.718,p=0.0494,consistency=0.70

Low-beta ITPC 13-20 Hz feature plot

feature note

輸出檔

Subject feature mapsnormal_left_right_feature_maps.csv
Feature summarynormal_left_right_feature_summary.csv
Analysis summarynormal_left_right_feature_exploration_summary.json
Feature notesnotes/