流程概述：IMU 姿勢分類（RNN-LSTM）

 資料格式（homework.csv，欄位包含 b1..b6 為 IMU 6軸數據，type 為動作類別）所設計的 LSTM 姿勢分類完整流程與步驟規範。這套流程可直接應用於姿勢分類模型訓練與驗證。

---

🧭 流程概述：IMU 姿勢分類（RNN-LSTM）

步驟	作業內容說明	工具/技術
Step 1	資料讀取與預處理	pandas, numpy
Step 2	資料標準化（Zero-mean, Unit-std）	sklearn.preprocessing.StandardScaler
Step 3	切割序列樣本（Sliding Window）	np.lib.stride_tricks.sliding_window_view 或手工實作
Step 4	建立 PyTorch Dataset / DataLoader	torch.utils.data.Dataset
Step 5	定義 LSTM 模型	torch.nn.LSTM
Step 6	模型訓練與驗證	CrossEntropyLoss, Adam optimizer
Step 7	驗證分類準確率與混淆矩陣	sklearn.metrics.confusion_matrix

---

📝 改寫：使用 homework.csv

📁 資料格式假設

csv
time,b1,b2,b3,b4,b5,b6,type
0.00,0.15,-0.02,...,1
0.01,0.16,-0.01,...,1
...

✅ 實作步驟

python
import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler
from torch.utils.data import Dataset, DataLoader
import torch
import torch.nn as nn

# Step 1: 讀取資料
df = pd.read_csv('homework.csv')
imu_data = df[['b1', 'b2', 'b3', 'b4', 'b5', 'b6']].values
labels = df['type'].values

# Step 2: 資料標準化
scaler = StandardScaler()
imu_data = scaler.fit_transform(imu_data)

# Step 3: Sliding window 切割
def create_sequences(data, labels, seq_len=100):
    sequences, seq_labels = [], []
    for i in range(len(data) - seq_len):
        seq = data[i:i+seq_len]
        label = labels[i+seq_len-1]  # 最後時間點標籤
        sequences.append(seq)
        seq_labels.append(label)
    return np.array(sequences), np.array(seq_labels)

X_seq, y_seq = create_sequences(imu_data, labels, seq_len=100)

# Step 4: PyTorch Dataset 定義
class IMUDataset(Dataset):
    def __init__(self, X, y):
        self.X = torch.tensor(X, dtype=torch.float32)
        self.y = torch.tensor(y, dtype=torch.long)

    def __len__(self):
        return len(self.X)

    def __getitem__(self, idx):
        return self.X[idx], self.y[idx]

# 切分 train/val
train_size = int(0.8 * len(X_seq))
train_dataset = IMUDataset(X_seq[:train_size], y_seq[:train_size])
val_dataset = IMUDataset(X_seq[train_size:], y_seq[train_size:])
train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=64)

---

🧠 LSTM 模型定義

python
class IMULSTMClassifier(nn.Module):
    def __init__(self, input_dim=6, hidden_dim=64, num_layers=2, num_classes=4):
        super().__init__()
        self.lstm = nn.LSTM(input_dim, hidden_dim, num_layers, batch_first=True)
        self.fc = nn.Linear(hidden_dim, num_classes)

    def forward(self, x):
        out, _ = self.lstm(x)
        out = self.fc(out[:, -1, :])
        return out

---

🔁 模型訓練與驗證

python
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = IMULSTMClassifier().to(device)
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)

for epoch in range(20):
    model.train()
    for X_batch, y_batch in train_loader:
        X_batch, y_batch = X_batch.to(device), y_batch.to(device)
        optimizer.zero_grad()
        output = model(X_batch)
        loss = criterion(output, y_batch)
        loss.backward()
        optimizer.step()

    print(f"Epoch {epoch+1} done.")

---

📊 驗證與混淆矩陣

python
from sklearn.metrics import accuracy_score, confusion_matrix
model.eval()
all_preds, all_labels = [], []

with torch.no_grad():
    for X_batch, y_batch in val_loader:
        X_batch = X_batch.to(device)
        preds = model(X_batch).argmax(dim=1).cpu().numpy()
        all_preds.extend(preds)
        all_labels.extend(y_batch.numpy())

acc = accuracy_score(all_labels, all_preds)
print(f"Validation Accuracy: {acc:.2f}")
print("Confusion Matrix:\n", confusion_matrix(all_labels, all_preds))

---

📌 小提醒

• 可根據實際需要改變 sequence_length、hidden_dim 或增減 Dropout, BatchNorm。

• 若你的 homework.csv 有其他欄位命名，請依實際名稱調整欄位選取。

• 可進一步導入 TensorBoard 或 wandb 做可視化監控。