Charles04Ekanem/Voice-Denoise-Nigeria

Working with a hybrid architecture + hybrid dataset approach to reduce noise in speech within Nigerian urban centres. This employs a CNN-BiLSTM-Attention-Residual (CBAR) model trained with a mix of clean speech from VoiceBank-DEMAND and real-world urban noise (e.g Generators, Traffic, Chatter, etc). # Voice-Denoise-Nigeria Working with a hybrid architecture + hybrid dataset approach to reduce noise in speech within Nigerian urban centres. This employs a custom built CNN-BiLSTM-Attention-Residual (CBAR) model trained with a mix of clean speech from VoiceBank-DEMAND and real-world urban noise (e.g Generators, Traffic, Chatter, etc).