👁️‍🗨️ 1. Introudction

📎 2. Related works

🗼 3. Proposed Method

• interspeech paper에서는 2spk fixed를 사용했지만 TASLP에서는 variable speaker 실험

IV. EEND-NAA for a Known Number of Speakers [interspeech]

🧱 EEND-EDA background

A. Non-Autoregressive Attractor Estimation

B. Iterative Refinement

V. EEND-NAA for an Unknown Number of Speakers

**⚠️ [**Problem] Known spk.

SSAD (Single Speaker Activity Detection) Module:

🧩 EEND-NAA-Overest (Over K-Menas ⇒ binary filtering layer)

🧩 EEND-NAA-2step

🧩 EEND-NAA-1step

📄 4. Experimental Setup

🧪 5. Experiments