You are on page 1of 21

SPEECH EMOTION RECOGNITION

USING MFCC AND SVM


AGENDA
Abstract

Introduction

Literature Survey

Problem Statement

Objectives

Proposed Methodology

Block Diagram

References

ABSTRACT

• In this we will implement speech emotion recognition


using SVM training and testing.
• There are 6 different types of speeches we classified.
• We used emotions such as : Anger, Disguise, Fear, Happy,
Sad, Surprise.

• We extract some features from speech such as pitch,


melfilterbank, MFCC.
Background
 “Speech” is the most natural way for people to
interact with others
 People can exchange information and emotions
rapidly through speech
 Speech signal can be received more easily through
mobile device and other media
 Recognizing the emotions from speech is
beneficial for many applications
 Human-machine interaction, diagnostic tool, call
center, entertainment, etc.
 Human emotion description
 Categorical description
 “Basic emotion” exists, if it is differentiable from all other
emotions [Ekman, 1999].
 Big six: Anger, Fear, Sadness, Joy, Disgust, Surprise.
 Dimensional description
 Arousal, Valence, Dominance.
 Category description is chosen in this work
 It can be easily used in many applications since the
meaning is more understandable and specific
 Database resource is more abundant
INTRODUCTION
 Database of Polish Emotional Speech comprises 240 recordings
from 8 actors (4 females and 4 males).
 Recordings for every speaker were made during a single session.
 Each speaker utters five different sentences with six types of
emotional load: joy, boredom, fear, anger, sadness and neutral ("no
emotion").
 Recently, the information provided by cameras and microphones
enable the computer to interact with the user though advanced
image and sound processing techniques.
 Research in automatic detection of expressed emotion is quite
limited. Recent research in this aspect mostly focuses on
classification, in the other words, mostly aims at ascertaining the
emotion of each utterance. This, however, is insufficient for our
applications.
MFCC (Mel-frequency cepstral coefficients )

 Mel-frequency cepstral
coefficients (MFCCs) are
coefficients that collectively
make up an MFC.
 They are derived from a type
of cepstral representation of
the audio clip (a nonlinear
"spectrum-of-a-spectrum").
LITERATURE SURVEY
 SVM(Support Vector Machine)

 SVM (Support Vector Machines) is popular strategies for


supervised machine learning and classification but at the cost of
increased memory and computing resources.
 For very high precision, you need a lot of memory since you have
to store all the support vectors in memory and this number grows
sort of linearly with the training dataset size.
LITURATURE
 The literature in speech emotion detection is not very rich
and researchers are still debating what features influence
the recognition of emotion in speech.
 There is also considerable uncertainty as to the best
algorithm for classifying emotion, and which emotions to
class together.
 The researchers usually deal with elicited and acted
emotions in a lab setting from few actors, just like in our
case.
 However, in the real problem, different individuals reveal
their emotions in a diverse degree and manner.
PROBLEM STATEMENT

 To design & implementation of speech emotion


recognition by MFCC and SVM classifier.
OBJECTIVES
 To study different existing methods for speech emotion
recognition .
 To implement a MATLAB code for speech emotion
recognition by MFCC and SVM classifier.
PROPOSED METHOD
TRAINING SPEECHES TESTING SPEECHE

READ SPEECHES READ SPEECH

SEGMENTED SPEECH SEGMENTED SPEECH

FEATURE EXTRACTION FEATURE EXTRACTION


1. PITCH 1. PITCH
2. MELFILTERBANK 2. MELFILTERBANK
3. MFCC 3. MFCC

SVM TRAINING SVM TESTING

MATCHED EMOTION
Fig : Block Diagram OF Proposed Work
Contd…
1. preprocessing
2. Segmentation: Image Segmentation
3.Feature Extraction: Extract Features by using
MFCC,PITCH,MELFILTERBANK.
4. Training: Training of SVM
5.Testing : Select a query Speech and Test(SVM
classifier)
Technology Used : MATLAB

MATLAB (matrix laboratory) is a multi-

paradigm numerical computing environment


and fourth-generation programming language.
Its one of the easiest and most productive

software development environment for


engineers.
MATLAB toolboxes Used

Image Processing Toolbox


Image Acquisition Toolbox
Computer Vision Toolbox
ADVANTAGES
Simple algorithm
Saves time
Problems such as local optima and sensitivity to initialization
are reduced.
SVM is having very high accuracy.
CONCLUSION
 Finally we proposed the method for detection of Brest
cancer using MATLAB successfully.
 The main purpose of this paper is to get perfect results for
emotion recognition using MFCC and pitch features.
 The proposed method is compared with SVM classifier.
Our experimental results indicate that the recognition rate
in the proposed method is reasonable for the EMOTION
recognition.
APPLICATIONS

Automated call centers


Assessing drivers mental stress.
Future work
 Emotional information is conveyed by multimodal cues,
including audio and visual expression. Thus, the multilevel
unit chunking and HCM will be conducted on audio-visual
data in the future
 Our approach will be applied to natural conversational
data, which might have more emotional transition and
change, to seek its effectiveness and possibilities for real
world application.
References

[1] M. Pantic and L. J. M. Rothkrantz, “Toward an affect-sensitive


multimodal human-computer interaction”,
[2] R. Cowie, E. Douglas-Cowie, N. Tsapatsoulis, G. Votsis, S. Kollias, W.
Fellenz, and J. G. Taylor, “Emotion recognition in humancomputer
interaction.”,
[3] J. Ang, R. Dhillon, A. Krupski, E. Shriberg, and A. Stolcke, “Prosody-
based automatic detection of annoyance and frustration in human–
computer dialog”,
. [4] Theodoros Iliou and Christos-Nikolaos Anagnostopoulos, "Statistical
Evaluation of Speech Features for Emotion Recognition,"
[5] L. C. De Silva and P. C. Ng, “Bimodal emotion recognition”, Proc. 4th
IEEE Intrnational Conference on Automatic Face and Gesture Recognition
(FG)

You might also like