Mixed Voice Dataset (4s) — Real/Fake Voice + Real/Fake Background

4-second mixed audio dataset for deepfake voice detection research. Combines real/fake human voice (ASVspoof 2021 DF eval set, ≥4s clips) with real/fake background sounds (EnvSDD). 4 categories × 3 volume levels × 1000 files = 12,000 WAV files (16kHz). Categories: real_voice_real_bg, real_voice_fake_bg, fake_voice_real_bg, fake_voice_fake_bg. Volume levels: balanced (60/60 dB), voice_dominant (80/60 dB), bg_dominant (60/80 dB).