Soheil Khorram

Title: Soheil Khorram: Innovator in Speech Recognition Technology

Introduction

Soheil Khorram is a prominent inventor based in Redwood City, California. He has made significant contributions to the field of speech recognition technology, holding three patents that showcase his innovative approach to machine learning and audio processing.

Latest Patents

One of his latest patents is titled "Contrastive Siamese Network for Semi-Supervised Speech Recognition." This method involves receiving a plurality of unlabeled audio samples corresponding to spoken utterances that are not paired with transcriptions. The process includes generating a sequence of encoder outputs and modifying time characteristics to create target branch outputs. Additionally, it performs augmentation on the unlabeled audio samples to generate predictions and determine an unsupervised loss term for updating the audio encoder parameters.

Another notable patent is the "Semi-Supervised Training Scheme for Speech Recognition." This method focuses on receiving a sequence of acoustic frames from unlabeled audio samples. It generates a target higher-order feature representation using a supervised audio encoder and augments the sequence of acoustic frames. The method also determines an unsupervised loss term based on the predicted higher-order feature representation, which is crucial for updating the speech recognition model parameters.

Career Highlights

Soheil Khorram is currently employed at Google Inc., where he continues to develop cutting-edge technologies in speech recognition. His work has significantly advanced the capabilities of machine learning applications in audio processing.

Collaborations

He has collaborated with notable coworkers, including Hasim Sak and Anshuman Tripathi, who are also recognized for their contributions to the field of speech recognition and machine learning.

Conclusion

Soheil Khorram's innovative work in speech recognition technology exemplifies the impact of modern inventions on communication. His patents reflect a deep understanding of machine learning and audio processing, positioning him as a key figure in the industry.