Speaker Separation Using Speaker Inventories and Estimated Speech

We propose speaker separation using inventories and estimated speech (SSUSIES), a framework leveraging profiles for separation. SSUSIES contains two methods, (SSUSI) (SSUES). SSUSI performs with the help of inventory. By combining advantages permutation invariant training (PIT) extraction, significantly outperforms conventional approaches. SSUES is widely applicable technique that can substantially improve performance output first-pass evaluate models on both recognition metrics.