Update README.md

README.md

@@ -5,10 +5,6 @@ language:
 model-index:
 - name: TinyMyo
   results:
-
-  # -------------------------
-  # Hand Gesture Classification
-  # -------------------------
   - task:
       type: gesture-classification
     dataset:
@@ -23,7 +19,6 @@ model-index:
       type: f1
       value: 0.7797
       verified: false
-
   - task:
       type: gesture-classification
     dataset:
@@ -38,7 +33,6 @@ model-index:
       type: f1
       value: 0.9674
       verified: false
-
   - task:
       type: gesture-classification
     dataset:
@@ -53,10 +47,6 @@ model-index:
       type: f1
       value: 0.9755
       verified: false
-
-  # -------------------------
-  # Generic Neuromotor Interface (Meta RL)
-  # -------------------------
   - task:
       type: gesture-classification
     dataset:
@@ -67,10 +57,6 @@ model-index:
       type: classification-error-rate
       value: 0.153
       verified: false
-
-  # -------------------------
-  # Hand Kinematic Regression
-  # -------------------------
   - task:
       type: kinematic-regression
     dataset:
@@ -89,10 +75,6 @@ model-index:
       type: r2
       value: 0.62
       verified: false
-
-  # -------------------------
-  # Silent Speech Synthesis
-  # -------------------------
   - task:
       type: speech-synthesis
     dataset:
@@ -103,10 +85,6 @@ model-index:
       type: word-error-rate
       value: 0.3354
       verified: false
-
-  # -------------------------
-  # Silent Speech Recognition
-  # -------------------------
   - task:
       type: speech-recognition
     dataset:
@@ -117,6 +95,10 @@ model-index:
       type: word-error-rate
       value: 0.3395
       verified: false
+tags:
+- emg
+- bio-signals
+- foundation-model
 ---
 
 <div align="center">
@@ -209,7 +191,7 @@ Unlike EEG FMs that mix channels early, TinyMyo uses **per-channel patching**:
 * Patch stride: **20 samples**
 * Tokens/channel: **50**
 * Total seq length: **800 tokens** (16 x 50)
-* Positional encoding: **RoPE (rotary)**
+* Positional encoding: **RoPE**
 
 This preserves electrode-specific structure while allowing attention to learn cross-channel relationships.
 
@@ -263,18 +245,17 @@ TinyMyo generalizes across **gesture classification**, **kinematic regression**,
 
 Evaluated on:
 
-* **Ninapro DB5** (52 classes, 10 subjects)
-* **EPN-612** (5 classes, 612 subjects)
-* **UCI EMG** (6 classes, 36 subjects)
-* **Meta Neuromotor Interface** (9 gestures)
+* **Ninapro DB5** (52 classes, 10 subjects, 200 Hz)
+* **EPN-612** (5 classes, 612 subjects, 200 Hz)
+* **UCI EMG** (6 classes, 36 subjects, 200 Hz)
 
 ### Preprocessing
 
 * EMG filtering: **20–90 Hz bandpass + 50 Hz notch**
 * Window sizes:
 
-  * **200 ms** (best for DB5)
-  * **1000 ms** (best for EPN, UCI)
+  * **200 samples** (1 sec, best for DB5)
+  * **1000 samples** (5 sec, best for EPN, UCI)
 
 ### Linear Classification Head
 
@@ -285,10 +266,9 @@ Evaluated on:
 
 | Dataset                  | Metric | Result            |
 | ------------------------ | ------ | ----------------- |
-| **Ninapro DB5** (200 ms) | Acc    | **89.41 ± 0.16%** |
-| **EPN-612** (1000 ms)    | Acc    | **96.74 ± 0.09%** |
-| **UCI EMG** (1000 ms)    | Acc    | **97.56 ± 0.32%** |
-| **Neuromotor**           | CLER   | **0.153 ± 0.006** |
+| **Ninapro DB5** (1 sec)  | Acc    | **89.41 ± 0.16%** |
+| **EPN-612** (5 sec)      | Acc    | **96.74 ± 0.09%** |
+| **UCI EMG** (5 sec)      | Acc    | **97.56 ± 0.32%** |
 
 TinyMyo achieves **new state-of-the-art** on DB5, EPN-612, and UCI.
 
@@ -297,7 +277,7 @@ TinyMyo achieves **new state-of-the-art** on DB5, EPN-612, and UCI.
 ## 4.2 Hand Kinematic Regression (Ninapro DB8)
 
 * Predict **5 joint angles**
-* Windows: **200 ms** or **1000 ms**
+* Windows: **100 ms** or **500 ms**
 * Normalization: z-score only
 
 ### Regression Head (~788k params)
@@ -309,7 +289,7 @@ TinyMyo achieves **new state-of-the-art** on DB5, EPN-612, and UCI.
 
 ### Performance
 
-* **MAE = 8.77 ± 0.12°** (1000 ms)
+* **MAE = 8.77 ± 0.12°** (500 ms)
 
 Note: Prior works reporting ~6.9° MAE are **subject-specific**; TinyMyo trains a **single cross-subject model**, a significantly harder setting.
 
@@ -354,7 +334,7 @@ TinyMyo runs efficiently on **GAP9 (RISC-V)** via:
 * Integer LayerNorm, GELU, softmax
 * Static memory arena via liveness analysis
 
-### Runtime (DB5 pipeline)
+### Runtime (EPN-612 dataset)
 
 * **Inference time**: **0.785 s**
 * **Energy**: **44.91 mJ**