The latest progress and development trend of 2022 intelligent voice technology

Learn in depth 、 Driven by big data and big computing power , Speech enhancement 、 Intelligent speech technology represented by recognition and synthesis has been applied in many applications . I've compiled some cutting-edge reports for you , Wen Wuke Free access .

No.1

New progress and development trend of intelligent voice technology

Speaker ： Xie Lei

Professor of Northwestern Polytechnic University , With concurrent

Head of the audio speech and Language Processing Laboratory of West University of Technology

Abstract ：

This report will combine the recent research results of the audio speech and language processing research group of Western Polytechnic University with the development status of intelligent speech technology , Focus on speech enhancement 、 Recent advances in recognition and synthesis . At the same time, with the continuous expansion of scenarios and Applications , Challenges of intelligent voice technology and prospects for future development .

No.2

Research progress of end-to-end sound source separation

Speaker ： Luo Yi

PhD student at Neural acoustic processing lab (Naplab),Columbia University.

Abstract ：

Recent progress in deep learning methods for the task of source separation have significantly advanced the state-of-the-art. 

Among all the recent proposals, end-to-end systems that take waveform as input and directly generate waveforms have shown their advantage on both the system performance and the flexibility. In this talk, I will briefly go through some of the recent advances in the problem of end-to-end neural source separation. I will start with the general problem definition of source separation, then introduce several single-channel and multi-channel approaches, and conclude with the challenges and future works in this area.

Scan the code to get all the reports for free
↓↓↓

No.3

Multi speaker segmentation and clustering based on deep learning

Speaker ： juck

University of Cambridge Research Associate

JD technical advisor  

Abstract ：

This open class first introduces the traditional multi speaker segmentation and clustering system of Cambridge University , The system has obtained ASRU 2015 MGB The champion of the speaker segmentation clustering task in the challenge , Then it introduces some work of the team recently using deep neural network to segment different parts of the clustering system . Finally, it also includes the discussion of some hot issues in the research of multi speaker segmentation and clustering , Including how to achieve a complete end-to-end neural network （ Trainable ） System and how to integrate segmentation and clustering with speech separation and recognition .

No.4

Sound event detection under weak tagging

Speaker ： Wang Yun  

Facebook The artificial intelligence application research group studies scientists

Carnegie Mellon University （CMU） Institute of computer technology （LTI） Doctor

Abstract ：

Sound event detection （sound event detection）, It refers to the detection of gunfire in the audio 、 Dog barking and other events , And mark their start and end time . Because it is troublesome to manually standard the start and end time for training data , Therefore, the actual training data is often only weakly labeled —— Only the event type contained in each sound is marked , But the starting and ending time is not marked . This lecture discusses how to use 「 Learn from various examples 」（multiple instance learning） Method , Using weak labeled data to train sound event detection system , The key is how to select the aggregate function , Maintain the balance between false detection and missed detection . The experience gained from this lecture , It can also be used for reference 「 Learn from various examples 」 In the task of .

No.5

Intelligent voice development status and data set introduction

Speaker ： Chen Guoguo

SEASALT.AI cofounder

Dr. Johns Hopkins University

Abstract ：

Share and discuss the current problems in the voice field , example ： When intelligent voice is landing on the embedded device , Compared with the server side, what factors need to be considered ; At the same time, combine their own scientific research and entrepreneurial experience to scientific research colleagues 、 Students in school 、 Some practical suggestions , Let's avoid detours ！

No.6

Research progress of accent and dialect speech recognition

Speaker ： Tangzhiyuan

Dr. Lian Pei, Chinese Academy of Sciences and Tsinghua University

Tsinghua postdoctoral

Abstract ：

Speech recognition technology has been widely used in daily life , However, its performance or experience in accent or dialect is still not satisfactory . This report gives a quick review of the research progress of accent and dialect speech recognition in recent years , And further introduces the data related to accent or dialect speech recognition 、 Benchmarks and competitions , And some feasible research directions .

Scan the code to get all the reports for free
↓↓↓