Auxiliary Networks for Joint Speaker Adaptation and Speaker Change Detection

Speaker adaptation and speaker change detection have both been studied extensively to improve automatic speech recognition (ASR). In many cases, these two problems are investigated separately: is implemented first obtain single-speaker regions, then performed using the derived segments for improved ASR. However, in an online setting, we want achieve goals a single pass. this study, propose neural network architecture that learns embedding from which it can perform ASR detection. The proposed computed self-attention based on auxiliary attached main network. by subtracting, activations, segment dependent affine transformation of learned embedding. experiments broadcast news dataset Switchboard conversational dataset, test our system utterances with point them show method achieves significantly better performance as compared unadapted (10-14% relative reduction word error rate (WER)). also outperforms three different segmentation methods followed (around 10% WER).