Tae Jin Park

Title: Innovations of Tae Jin Park in Acoustic Modeling

Introduction

Tae Jin Park is an accomplished inventor based in Los Angeles, CA. He has made significant contributions to the field of speech processing through his innovative patent. His work focuses on utilizing deep neural networks to enhance acoustic modeling techniques.

Latest Patents

Tae Jin Park holds a patent for "Deep multi-channel acoustic modeling using frequency aligned network." This patent describes techniques for speech processing that leverage a deep neural network (DNN) based acoustic model front-end. The new modeling approach directly addresses multi-channel audio data received from a microphone array. It employs a first model, which includes a frequency aligned network (FAN) architecture, to perform spatial filtering. This process generates a first feature vector by processing individual frequency bins separately, ensuring that multiple frequency bins are not combined. The first feature vector can be utilized similarly to beamformed features generated by an acoustic beamformer. Subsequently, a second model processes the first feature vector to create a lower-dimensional representation, while a third model performs acoustic unit classification to generate text data. This DNN front-end significantly improves performance, even with a reduction in the number of microphones.

Career Highlights

Tae Jin Park is currently employed at Amazon Technologies, Inc., where he continues to innovate in the field of acoustic modeling. His work has garnered attention for its potential applications in various speech processing technologies.

Collaborations

Tae Jin Park has collaborated with notable colleagues, including Minhua Wu and Shiva Kumar Sundaram. Their combined expertise contributes to the advancement of technologies in their field.

Conclusion

Tae Jin Park's innovative work in deep multi-channel acoustic modeling showcases his expertise and commitment to advancing speech processing technologies. His contributions are paving the way for improved performance in various applications.