Kaizhi Qian

Title: Kaizhi Qian: Innovator in Speech Recognition Technology

Introduction

Kaizhi Qian is a prominent inventor based in Champaign, IL (US). He has made significant contributions to the field of speech recognition technology, holding a total of 7 patents. His work focuses on advancing machine learning techniques to improve audio processing and understanding.

Latest Patents

One of his latest patents is titled "Self-supervised speech recognition." In this innovation, one or more computer processors obtain an initial subnetwork at a target sparsity and an initial pruning mask from a pre-trained self-supervised learning (SSL) speech model. The processors then finetune the initial subnetwork by zeroing out one or more masked weights specified by the initial pruning mask. They train a new subnetwork from the zeroed-out subnetwork and prune weights of the lowest magnitude to satisfy the target sparsity. Finally, the processors classify an audio segment with the finetuned subnetwork.

Another notable patent is "Global prosody style transfer without text transcriptions." This method utilizes a machine learning model for disentanglement of prosody in spoken natural language. The process includes encoding spoken natural language to produce content code, resampling the content code to obscure the prosody without text transcriptions, and decoding the prosody-obscured content code to synthesize speech indirectly based on the content code.

Career Highlights

Kaizhi Qian has worked with notable companies such as IBM and Logos Care, Inc. His experience in these organizations has allowed him to develop and refine his innovative ideas in speech recognition technology.

Collaborations

Throughout his career, Kaizhi has collaborated with talented individuals, including Yang Zhang and Chuang Gan. These partnerships have contributed to the advancement of his research and inventions.

Conclusion

Kaizhi Qian is a distinguished inventor whose work in speech recognition technology has the potential to transform how we interact with machines. His innovative patents reflect his commitment to enhancing audio processing through advanced machine learning techniques.