Speech to Speech Synthesis for Voice Impersonation

Numerous models have shown great success in the fields of speech recognition as well as speech synthesis, but models for speech to speech processing have not been heavily explored. We propose Speech to Speech Synthesis Network (STSSN), a model based on current state of the art systems that fuses the two disciplines in order to perform effective speech to speech style transfer for the purpose of voice impersonation. We show that our proposed model is quite powerful, and succeeds in generating realistic audio samples despite a number of drawbacks in its capacity. We benchmark our proposed model by comparing it with a generative adversarial model which accomplishes a similar task, and show that ours produces more convincing results.

Markerless Suture Needle 6D Pose Tracking with Robust Uncertainty Estimation for Autonomous Minimally Invasive Robotic Surgery

Suture needle localization plays a crucial role towards autonomous suturing. To track the 6D pose of a suture needle robustly, previous approaches usually add markers on the needle or perform complex operations for feature extraction, making these methods difficult to be applicable to real-world environments. Therefore in this work, we present a novel approach for markerless suture needle pose tracking using Bayesian filters. A data-efficient feature point detector is trained to extract the feature points on the needle. Then based on these detections, we propose a novel observation model that measures the overlap between the detections and the expected projection of the needle, which can be calculated efficiently. In addition, for the proposed method, we derive the approximation for the covariance of the observation noise, making this model more robust to the uncertainty in the detections. The experimental results in simulation show that the proposed observation model achieves low tracking errors of approximately 1.5mm in position in space and 1 degree in orientation. We also demonstrate the qualitative results of our trained markerless feature detector combined with the proposed observation model in real-world environments. The results show high consistency between the projection of the tracked pose and that of the real pose.