A Study to Assess the Effectiveness of Planned Teaching Programme on Knowledge Regarding Maintaining Airway Patency in Patients with Mechanical Ventilator

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 6 Issue 1, November-December 2021 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD47958 | Volume – 6 | Issue – 1 | Nov-Dec 2021 Page 922
Speech Emotion Recognition Using Neural Networks
Anirban Chakraborty
Research Scholar, Department of Artificial Intelligence, Lovely Professional University, Jalandhar, Punjab, India
ABSTRACT
Speech is the most natural and easy method for people to
communicate, and interpreting speech is one of the most
sophisticated tasks that the human brain conducts. The goal of
Speech Emotion Recognition (SER) is to identify human emotion
from speech. This is due to the fact that tone and pitch of the voice
frequently reflect underlying emotions. Librosa was used to analyse
audio and music, sound file was used to read and write sampled
sound file formats, and sklearn was used to create the model. The
current study looked on the effectiveness of Convolutional Neural
Networks (CNN) in recognising spoken emotions. The networks'
input characteristics are spectrograms of voice samples. Mel-
Frequency Cepstral Coefficients (MFCC) are used to extract
characteristics from audio. Our own voice dataset is utilised to train
and test our algorithms. The emotions of the speech (happy, sad,
angry, neutral, shocked, disgusted) will be determined based on the
evaluation.
KEYWORDS: Speech emotion, Energy, Pitch, Librosa, Sklearn,
Sound file, CNN, Spectrogram, MFCC
How to cite this paper: Anirban
Chakraborty "Speech Emotion
Recognition Using Neural Networks"
Published in
International
Journal of Trend in
Scientific Research
and Development
(ijtsrd), ISSN: 2456-
6470, Volume-6 |
Issue-1, December
2021, pp.922-927, URL:
www.ijtsrd.com/papers/ijtsrd47958.pdf
Copyright © 2021 by author(s) and
International Journal of Trend in
Scientific Research and Development
Journal. This is an
Open Access article
distributed under the
terms of the Creative Commons
Attribution License (CC BY 4.0)
(http://creativecommons.org/licenses/by/4.0)
I. INTRODUCTION
Speech emotion recognition (SER) is a technique that
extracts emotional features from speech by analysing
distinctive characteristics and the acquired emotional
change. At the moment, voice emotion recognition is
a developing artificial intelligence cross-field [1]. A
voice emotion processing and recognition system is
made up of three parts: speech signal acquisition,
feature extraction, and emotion recognition. In this
method, the extraction quality has a direct impact on
the accuracy of speech emotion identification. In
feature extraction, the entire emotion sentence was
frequently used as a unit for feature extraction and
extraction contents. The neural networks of the
human brain are highly capable of learning high-level
abstract notions from low-level information acquired
by the sensory periphery. Humans communicate
through voice, and interpreting speech is one of the
most sophisticated operations that the human brain
conducts. It has been argued that children who are not
able to understand the emotional states of the
speakers developed poor social skills and in some
cases they show psychopathological symptoms [2, 3].
This highlights the importance of recognizing the
emotional states of speech in effective
communication. Detection of emotion from facial
expressions and biological measurements such as
heart beats or skin resistance formed the preliminary
framework of research in emotion recognition[4].
More recently, emotion recognition from speech
signal has received growing attention. The traditional
approach toward this problem was based on the fact
that there are relationships between acoustic features
and emotion. In other words, the emotion is encoded
by acoustic and prosodic correlates of speech signals
such as speaking rate, intonation, energy, formant
frequencies, fundamental frequency (pitch), intensity
(loudness), duration (length), and spectral
characteristic (timbre) [5, 6]. There are a variety of
machine learning algorithms that have been examined
to classify emotions based on their acoustic correlates
in speech utterances. In the current study, we
investigated the capability of convolutional neural
networks in classifying speech emotions using our
own dataset. There are a variety of machine learning
algorithms that have been examined to classify
emotions based on their acoustic correlates in speech
utterances. In the current study, we investigated the
capability of convolutional neural networks in
classifying speech emotions using our own dataset.
The specific contribution of this study is using
IJTSRD47958

International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470
wideband spectrograms instead of narrow-band
spectrograms as well as assessing the effect of data
augmentation on the accuracy of models. Our results
revealed that wide-band spectrograms and data
augmentation equipped CNNs to achieve the state-of-
the art accuracy and surpass human performance.
Fig.1. Speech emotion recognition block diagram
II. RELATED WORK
Most of the papers published in last decade use spectral and prosodic features extracted from raw audio signals.
The process of emotion recognition from speech involves extracting the characteristics from a corpus of
emotional speech selected or implemented, and after that, the classification of emotions is done on the basis of
the extracted characteristics. The performance of the classification of emotions strongly depends on the good
extraction of the characteristics (such as combination of MFCC acoustic feature with the energy prosodic feature
[7]. Yixiong Pan in [8] used SVM for three class emotion classification on Berlin Database of Emotional Speech
[9] and achieved 95.1% accuracy.
Norooziet.al. Proposed a versatile emotion recognition system based on the analysis of visual and auditory
signals. He used 88 features (Mel frequency cepstral coefficients
(MFCC), filter bank energies (FBEs)) using the Principal Component Analysis (PCA) infeature extraction to
reduce the dimension of features previously extracted revealed that wide-band spectrograms and data
augmentation equipped CNNs to achieve the state-of-the art accuracy and surpass human performance.
The performance of the classification of emotions strongly depends on the good extraction of the characteristics
(such as combination of MFCC acoustic feature with the energy prosodic feature [7]. Yixiong Pan in [8] used
SVM for three class emotion classification on Berlin Database of Emotional Speech [9] and achieved 95.1%
accuracy.
Norooziet.al. proposed a versatile emotion recognition system based on the analysis of visual and auditory
signals. He used 88 features (Mel frequency cepstral coefficients (MFCC), filter bank energies (FBEs)) using the
Principal Component Analysis (PCA) in feature extraction to reduce the dimension of features previously
extracted [10]. S. Lalitha in [11] used pitch and prosody features and SVM classifier reporting 81.1% accuracy
on 7 classes of the whole Berlin Database of Emotional Speech. Zamil et al also used the spectral characteristics
which is the 13 MFCC obtained from the audio data in their proposed system to classify the 7 emotions with the
Logistic Model Tree (LMT) algorithm with an accuracy rate 70% [12]. Yu zhou in [13] combined prosodic and
spectral features and used Gaussian mixture model super vector based SVM and reported 88.35% accuracy on 5
classes of Chinese-LDC corpus.
H.M Fayek in [14] explored various DNN architecture and reported accuracy around 60% on two different
database eNTERFACE [15] and SAVEE [16] with 6 and 7 classes respectively. Fei Wang used combination of
Deep Auto Encoder, various features and SVM in [17] and reported 83.5% accuracy on 6 classes of Chinese
emotion corpus CASIA. In contrast to these traditional approaches more novel papers have been published
recently employing Deep Neural Networks into their experiments with the promising results. Many authors
agree that the most important audio characteristics to recognize emotions are spectral energydistribution, Teager
Energy Operator (TEO) [18], MFCC, Zero Crossing Rate (ZCR), and the energy parameters of the filter bank
energies (FBEs) [19].
III. TRADITIONAL SYSTEM
The traditional system was based on the analysis and comparison of all kinds of emotional characteristic
parameters, selecting emotional characteristics with high emotional resolution for feature extraction. In general,
the traditional emotional feature extraction concentrates on the analysis of the emotional features in the speech
from time construction, amplitude construction, and fundamental frequency construction and signal feature [28].

IV. PROPOSED METHOD
Convolutional Neural Network (CNN) is used to classify the emotions (happy, sad, angry, neutral, surprised,
disgust) and to predict the output by showing its accuracy.
The given speech is plotted as spectrogram by using matplot library and this is used as input for CNN to build
the model.
Fig.2. Flow diagram of proposed system
A. Data Set Collection
The first step is to create an empty dataset that will hold the training data for the model. After creating an empty
dataset, the data’s (audio) have to be recorded and labeled in different classes. Once the labeling is done, the
data’s have to be preprocessed which will produce the clear pitch of the data by removing its unwanted
background noise. After preprocessing the data’s are classified into train dataset and test dataset, where the train
dataset hold 75% of the data and the test dataset holds 25% of the data.
B. Feature Extraction of Speech Emotion
Human speech consists of many parameters which show the emotions compromise in it. As there is change in
emotions these parameters also gets changed. Hence it’s necessary to select proper feature vector to identify the
emotions. Features are categorized as excitation source features, spectral features, and prosodic features.
Excitation source features are achieved by suppressing characteristics of vocal tract (VT). Spectral features used
for emotion recognition are linear prediction coefficients (LPC), Perceptual Linear prediction coefficients
(PLPCs), Mel-frequency cepstral coefficients (MFCC), linear prediction cepstrum coefficients (LPCC), and
perceptual linear prediction (PLP). The accuracy of differentiating different emotions can be achieved by using
MFCC, LFPC
[20, 21].
C. Mel-Frequency Cepstral Coefficients
The Mel-Frequency Cepstral Coefficients (MFCC) feature extraction method is a leading approach for speech
feature extraction. The various steps involved in MFCC feature extraction are:
Fig.3. Flow of MFCC A/D conversion:
This converts the analog signal into discrete space.
Pre-emphasis:
This boosts the amount of energy in the high frequencies.

Windowing:
Windowing involves the slicing of audio waveform into sliding frames. Discrete Fourier Transform:
DFT is used to extract information in the frequency domain [22, 23].
D. Classifiers
After extracting features of speech, it is essential to select a proper classifier. Classifiers are used to classify
emotions. In the current study, we use Convolutional Neural Network (CNN). The term Convolutional comes
from the fact that Convolution-the mathematical operation is employed in these networks. Convolutional Neural
Networks is one of the most popular Deep Learning Models that have manifested remarkable success in the
research areas. CNN is a deep learning algorithm that takes image as an input, assign importance to various
aspects in the image and will be able to differentiate from other. Generally CNNs have three building blocks: the
convolutional layer, the pooling layer, and the fully connected layer. Following, we describe these building
blocks along with some basic concept such as soft max unit, rectified linear unit, and drop out.
Input layer: This layer holds the raw input image.
Convolution Layer: This layer computes the output volume by computing dot product between all filters
and image patch.
Activation Function Layer: This layer will apply element wise activation function to the output of
convolution layer.
Pool Layer: This layer is periodically inserted in CNN and its main function is to reduce the size of volume
which makes computation fast and reduces memory. The two types are Maxpooling and average pooling.
Fully-Connected Layer: This layer takes input from the previous layer and computes the class scores and
outputs the 1-D array of size equal to the number of classes [24, 25].
Fig.4. CNN Algorithm
V. APPLICATION
The applications of speech emotion recognition
system are, psychiatric diagnosis, conversation with
robots, intelligent toys, mobile based emotion
recognition, emotion recognition in call centre where
emotions of customer can be identified and can help
to get better service quality, intelligent tutoring
system, lie detection, games[26,27]. It is also used in
healthcare, Psychology, cognitive science and
marketing, voice-based virtual assistants.
VI. CONCLUSION
In this research, we suggested a technique for
extracting the emotional characteristic parameter
from an emotional speech signal using the CNN
algorithm, one of the Deep Learning methods.
Previous research relied heavily on narrow-band
spectrograms, which offer better frequency resolution
than wide-band spectrograms and can discern
individual harmonics. Wide-band spectrograms, on
the other hand, offer better temporal resolution than
narrow-band spectrograms and reveal distinct glottal
pulses that are connected with basic frequency and
pitch. On training data, CNNs perform admirably.
The current study's findings demonstrated CNNs'
ability to learn the fundamental emotional properties
of speech signals from their low-level representation
utilising wide-band spectrums.
VII. FUTURE SCOPE
For future work, we suggest to use audio-visual
database or audio-visual-linguistic databases to train
Deep Learning models where facial expressions and
semantic information are taken into account as well as
speech signals, which allows improving the
recognition rate of each emotion. In future, we can
think about using other types of features and apply
our system on other bases that are larger and used
other method for feature extraction.
REFERENCES
[1] Z. Yongzhao and C. Peng, “Research and
implementation of emotional feature extraction

and recognition in speech signal,” Journal of
Jiangsu University, volume. 26, No. 1, pp.72-
75, 2005.
[2] Monita Chatterjee, Danielle J Zion, Mickael L
Deroche, Brooke A Burianek, Charles J Limb,
Alison P Goren, Aditya M Kulkarni, and Julie
A Christensen. Voice emotion recognition by
Cochlear-implanted children and their
normally-hearing peers. Hearing research,
322:151-162, 2015.
[3] Nancy Eisenberg, Tracy L Spinrad, and Natalie
D Eggum. Emotional-related self-regulation
and its relation to children’s maladjustment.
Annual review of clinical pschycology, 6:495-
525, 2010.
[4] Harold Schlosberg. Three dimensions of
emotion. Psychological review, 61(2):81, 1954.
[5] Louis Ten Bosch. Emotions, speech and the asr
framework. Speech communication, 40(1-
2):213-225, 2003.
[6] Thomas S Polzin and Alex Waibel. Detecting
emotions in speech. In proceedings of the
CMC, volume 16. Citeseer, 1998.
[7] SurajTripathi, Abhay Kumar, Abhiram
Ramesh, Chirag Singh, PromodYenigalla,
“Speech emotion recognition using kernel
sparse representation based classifier,” in 2016
24th European Signal Processing
Conference(EUSIPCO), pp.374-377, 2016.
[8] Pan, Y., Shen, P. and Shen, L., 2012. Speech
emotion recognition using support vector
machine. International Journal of Smart Home,
6(2), pp.101-108.
[9] Burkhardt, F.,Taeschke, A.,Rolfes,
M.,Sendlmeier, W.F.andWeiss,B., 2015,
September. A database of German emotional
speech. In Interspeech (vol.5, pp.1517-1520).
[10] Noroozi, F., Marjavonic, M., Njegus, A.,
Escalera, S., &Anbarjafari, G. Audio-visual
emotion recognition in video clips. IEEE
Transactions on Affective Computing, 2017.
[11] Lalitha, S., Madhavan, A., Bhushan, B, and
Saketh, S., 2014, October. Speech emotion
recognition. In Advances in Electronics,
Computer characteristics. The performance of
the classification of emotions strongly depends
on the good extraction of the characteristics
(such as combination of MFCC acoustic feature
with the energy prosodic feature [7]. Yixiong
Pan in [8] used SVM for three class emotion
classification on Berlin Database of Emotional
Speech [9] and achieved 95.1% accuracy.
Noroozi et al proposed a versatile and
Communications (ICAECC), 2014
International Conference on (pp. 1-4). IEEE.
[12] Zamil, AdibAshfaq A., et al. “Emotion
Detection from Speech Signals using Voting
Machanism on Classified Frames.” 2019
International Conference on Robotics,
Electrical and Signal Processing Techniques
(ICREST). IEEE, 2019.
[13] Zhou, Y., Sun, Y., Zhang, J., and Yan, Y.,
2009, December. Speech emotion recognition
using both spectral and prosodic features. In
2009 International Conference on Information
Engineering and Computer Science (pp. 1-4).
IEEE.
[14] Fayek, H. M., M. Lech, and L. Cavedon.
“Towards real-time speech emotion recognition
using Deep Neural Networks.” Signal
Processing and Communication Systems
(ICSPCS), 2015 9th International Conference
on IEEE, 2015.
[15] Martin, O., Kotsia, L.,Macq, B., and Pitas, I.,
2006, April. The eNtERFACE’05 audio-visual
database. In 22nd International Conference on
Data Engineering Workshops (ICDEW’06) (pp
8-8). IEEE.
[16] Sanjitha. B. R, Nipunika. A, Rohita Desai.
“Speech Emotion Recognition using MLP”,
IJESC.
[17] Ray Kurzweil. The singularity is near. Gerald
Duckworth & Co, 2010.
[18] HadhamiAouani and Yassine Ben Ayed,
“Speech Emotion Recognition with Deep
Learning”, 24th International Conference on
Knowledge-Based and Intelligent Information
& Engineering Systems.
[19] PavolHarar, RadimBurget and Malay Kishore
Dutta “Efficiency of chosen ` speech
descriptors in relation to emotion recognition,”
EURASIP Journal on Audio, Speech, and
Music Processing, 2017.
[20] Idris I., Salam M.S. “Improved Speech
Emotion Classification from Spectral
Coefficient Optimization”. Lecture Notes in
Electrical Engineering, vol 387. Springer, 2016.
[21] Pao TL., Chen YT., Yeh JH., Cheng YM.,
Chien C.S. “Feature Combination for Better
Differentiating Anger from Neutral in
Mandarin Emotional Speech”, LNCS: Vol.
4738 Berlin: Springer 2007.

[22] J&M: Daniel Jurafsky and James H. Martin
(2008). Speech and Language Processing,
Pearson Education (2nd edition).
[23] HyenkHermansky, “Perceptual linear
Predictive (PLP) analysis of speech,” The
Journal of the Acoustical Society of America,
Vol.87, No.4, pp.1737-1752, 1980.
[24] A. Berg, J. Deng, and L. Fei-Fei. Large scale
visual recognition challenge 2010.
www.upgrad.com/blog/.

A Study to Assess the Effectiveness of Planned Teaching Programme on Knowledge Regarding Maintaining Airway Patency in Patients with Mechanical Ventilator

More Related Content

Similar to A Study to Assess the Effectiveness of Planned Teaching Programme on Knowledge Regarding Maintaining Airway Patency in Patients with Mechanical Ventilator (20)

More from ijtsrd (20)

Recently uploaded (20)

A Study to Assess the Effectiveness of Planned Teaching Programme on Knowledge Regarding Maintaining Airway Patency in Patients with Mechanical Ventilator