Font Size: a A A

Expression Recognition Based On Deep Learning And Its Application In Central Bank Communication

Posted on:2021-06-02Degree:MasterType:Thesis
Country:ChinaCandidate:B PanFull Text:PDF
GTID:2518306302976209Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
The rapid development of communication technology has expanded the channels of information dissemination in modern society,from newspapers and telephones to later video chats and live broadcasts.At the same time,it has also greatly changed the way governments communicate with financial markets.Effective communication between the central bank and the market can reduce information asymmetry between the market and the central bank,increase public trust in the central bank,and promote monetary policy goals.In the current monetary policy formulation process,increasing communication with the public and increasing policy selectivity has become a common consensus among countries.Whether the influence of the central bank on the financial market is conducive to economy is the object of many scholars' research.In modern society,the market can obtain the monetary policy and related information announced by the central bank through a variety of channels.However,the existing literature research mainly focuses on the interest rate adjustment,meeting type and text content issued by the central bank.Traditional research is unable to handle with new-type information transmission channels.In 2011,the Fed first held press conferences,in which the chairman can directly interpret policies and express personal opinions to the market.Under the high frequency communication,the influences of the chairman's emotion to the financial market presents new challenges to related research.The nature of central bank communication is a channel for transmitting information.The chairman is the central bank's agent,who will cause the public to judge the economic statements.However,few studies at home and abroad have focused on the impact of the emotion of chairman on the market.Social psychology believes that during communication,most of the impressions come from facial expressions and physical movements.Therefore,the incremental information contained in chairman's factors such as expression,voice,and speech cannot be underestimated.Based on this,the following work has been done in this article:First of all,this paper uses deep learning and computer vision technology to capture the chairman's facial emotions.With the rise of deep learning,neural networks have become the most important emotion recognition model.Res Net18 is chosen as the model,which is a classic model commonly used in computer vision research.This paper adopts Affect Net as training dataset,the largest existing facial emotion dataset,which collects and manually labels 440,000 facial pictures.The emotion recognition model has high accuracy and meet the research needs of this paper.Secondly,this paper gets FOMC press conference videos on the Fed's website as a source of chairman's mood data.During the press conference,the Fed chairman announced the results of the board vote,interpreted the upcoming policies,and answered questions from reporters.Each facial image fully reflects the expression of the chairman,with the results of positive emotion,valence and arousal.Last but not least,based on the previous two parts of the study,this paper conducts the empirical analysis of chairman communication emotion on financial markets.US bond market and stock market are selected as the representative of the financial market.The monetary policy information is decomposed into target surprise and path surprise as control variables for empirical research.Above all,this paper attempts to obtain the chairman 's facial image from press conference.Based on deep learning technology,the facial emotions are extracted and analyzed.This paper finds that:(1)The communication emotion of the chairman only has a short-term impact on the market,and there is no long-term impact.(2)Financial markets will react strongly to the negative emotions of the chairman,but will not respond to positive emotions.(3)The chairman 's communication emotion in Q&A contains more information.This article expands the content of traditional research and provides suggestions for concerned problems.
Keywords/Search Tags:Central Bank Communication, Facial Expression Recognition, Emotion, Monetary Policy
PDF Full Text Request
Related items