Bonfring International Journal of Research in Communication Engineering, Vol.

6, Special Issue, November 2016 48

Artificial Intelligence and its Application in Speech

Recognition
S.V. Viraktamath, Chaitra P. Shet and Pooja R. Nayak

Abstract--- Artificial intelligence involves studying the 2) Replaces human in intelligent tasks: Results in
thought processes of human beings and also representing building of system in order to help humans think
those processes via computers and robots like man made better, faster and deeper.
machines. Speech recognition system lets user do other works 3) Enhance human intelligence: It implies building
simultaneously so that the user can concentrate on program to exceed human intelligence.
observation and manual operations. With the help of speaker 4) Deals with Coherent discourse: Communication with
recognition technology we can help physically challenged people using natural language involves intelligent
skilled persons. So that they can do their works without the dialogue.
help of others. This Artificial Speech Recognition technology
is also used in various applications. Now days this technology IV. SPEAKER INDEPENDENCY
is also used by CID officers in order to trap the criminal Usually the speech quality varies from one person to
activities.This technology is also used in military applications. another person. So it becomes difficult to design an electronic
Keywords--- NLP, LPC, DFT, MMSE Technique. machine that recognizes everyone’s voice. The system is made
simpler and also more reliable by designing it in order to
recognize single person’s voice. The computer is trained to the
I. INTRODUCTION voice of the particular individual. Such developed system is
called speaker-dependent system. Speaker independent
A RTIFICIAL intelligence involves two basically deals
with studying the thought processes of human beings and
then representing those processes via computer[1].
systems can be used by anybody, as it recognizes any voice,
although the characteristics vary widely from one speaker to
There is one artificial intelligence method through which we another. These speaker independent systems are costly and
can communicate with a computer in a natural language like complex to construct. These systems have got limited
English which is termed as Natural Language vocabularies. There are some factors that may affect the
Processing(NLP). The main objective of a NLP program is to quality of speech recognition. That includes grammar used by
understand the applied input and then initiate related action. the speaker and accepted by the system, noise level, noise
type, position of the microphone, and speed and manner of the
II. DEFINITION user’s speech and so on.
Artificial Intelligence is the science and engineering of
V. ENVIRONMENTAL IMPACT
making intelligent machines, especially intelligent computer
programs so that it can serve most of user needs[2].AI implies The recognition rate usually drops widely when a system is
Artificial Intelligence. Intelligence is the factor which cannot trained and tested under different conditions. It is necessary
be defined but whereas AI can be defined as branch of that we should be aware of the variations present when
computer science which deals with the simulation of machine different microphones are used in training, testing, and also
that exhibits intelligent behavior as the human being. Speech during development of required procedures. Such that with
recognition is considered to be one of the applications of the this the accuracy of recognition systems can be improved to
artificial intelligence which mainly deals with the translation greater extent.
of user spoken words into the corresponding text. Accuracy of recognition systems are going to be degraded
mainly because of Acoustical distortions. Obstacles for
III. OBJECTIVES robustness include additive noise from machinery, competing
1) Deals with understanding human thinking capabilities: talkers, reverberation from surface reflections in a room, and
This means to obtain deep knowledge of human spectral shaping by microphones and also the vocal tracts of
memory, Problem solving ability, learning and individual speakers and so on. The sources of distortions are
decision making etc. categorized as additive noise and distortions resulting from the
convolution of the speech signal with an unknown linear
system. There are various algorithms proposed for speech
S.V. Viraktamath.
enhancement. They are given as follows:
Chaitra Prakash Shet, Student, Department of Electronics &
Communication, Shri Dharmasthala Manjunatheshwara College of 1. Spectral subtraction of the obtained DFT coefficients.
Engineering and Technology, India. E-mail:[email protected]
Pooja Ravindra Nayak, Student, Department of Electronics &
2. The DFT coefficients of corrupted speech are
Communication, Shri Dharmasthala Manjunatheshwara College of estimated by applying MMSE technique Convoluted
Engineering and Technology, India. E-mail:[email protected]
DOI:10.9756/BIJRCE.8199

ISSN 2277 - 5080 | © 2016 Bonfring

Bonfring International Journal of Research in Communication Engineering, Vol. 6, Special Issue, November 2016 49

Distortions are compensated by spectral equalization (LPC)[3].

method. The reconstruction of spectral envelop from a truncated set
3. Spectral subtraction and spectral equalization
of spectral coefficients is found to be much smoother than one
algorithm. which is obtained from LPC coefficients. Hence, it provides a
4. These methods are relatively successful, and all these more stable representation of a particular speaker’s utterances.
methods depend on the assumption of independence of
In order to represent the spectral dynamics typically the first
the spectral estimates across frequencies. MMSE and second order coefficients are extracted at every frame
estimator is used in order to get better performance in period. These coefficients which are derivatives of the time
which correlation among frequencies is modeled
function of the spectral coefficients are called the delta and
explicitly. delta-delta-spectral coefficients respectively.
VI. SPEAKER RELATED FEATURES VII. SPEECH IDENTIFICATION
Physiological and behavioral characteristics of the speaker Microphone is found to be the input device through which
are mainly considered as speaker identity. These features are
the user communicates with the application. For the speech
found both in the vocal tract characteristics and in the voice processing the Recognizer converts the analog signal into
source characteristics, as also in the dynamic features digital signal. As a result of which stream of text is
spanning the several segments. The most common short-term
generated[4].This source-language text acts as the input to the
spectral measurements currently used are the spectral Translation Engine, which converts it to the target language
coefficients derived from the Linear Predictive Coding text [5].

Figure 1: Speech Recognition

Salient Features 5. contextual selection of appropriate synonym
using Online thesaurus
1) Modes of applying input
6. Word addition, grammar creation and updating
a) Using Speech Engine
facility through online
b) Using soft copy
7. Includes Personal account creation and also inbox
2) Feature concerned with Interactive Graphical User
Management.
Interface
3) Format Retention
4) Standard and quick translation VIII. VOICE IDENTIFICATION
5) Comprising of various Interactive Pre-processing tool The field of computer science is made to design computer
1. Spell checker. systems that can recognize spoken words by the user. Voice
2. Phrase marker recognition implies identifying particular voice. A number of
3. Proper noun, date and other package specific voice recognition systems are available on the market most of
identifier Input Format which require a training session during which the computer
4. Input Format : txt, .doc .rtf Selection of multiple system is trained to identify particular voice and accent. Such
output which is user friendly systems are known as speaker dependent.

ISSN 2277 - 5080 | © 2016 Bonfring

Bonfring International Journal of Research in Communication Engineering, Vol. 6, Special Issue, November 2016 50

In most of the discrete speech systems it requires that the with a short pause so that identification becomes easier.
speaker speak slowly and distinctly and separate each word

BPF ADC
Digitized speech

BPF ADC I
N
P
U Template
T
BPF ADC

Search and pattern

matching program
BPF ADC

Output circuits
Circuits CPU

Figure 2: Voice Recognition

text to speech technologies. After voice is recognized
IX. VOICE PROCESSING processing of voice is done.In order to facilitate voice
The handling of voice through computer involves, storing processing by the system analog signal is converted to digital
of voice,forwarding it, voice response, voice recognition and signal

Display

Application
Dictating
Speaker
Speaker recognition
device
Commands to
computer

Input to other
devices

NLP Understanding

Figure 3: Voice Processing

to use their hands for this purpose[7].
X. APPLICATIONS
A radiologist usually scans hundreds of X-rays, ultra
Speech recognition has got many applications among sonograms, CT scans and simultaneously dictating
which it lets user do other works simultaneously. So that user conclusions to a speech recognition system which is connected
can concentrate on observation and manual operations, and to word processors [8].With the help of this radiologist can
still control the machinery by voice input commands. In the focus his attention on the images rather than writing the text.
field of military also speech recognition has got its application For making airline and hotel reservations also voice
[6]. The best example for reliable speech recognition recognition is used. [9]. With the help of this application a
equipment is Voice control of weapons, computers take input user requires simply stating his needs, in order to make
as simply spoken words for example the commands given by reservation, cancel a reservation, or making any enquiries
the pilots through their microphones such that they don’t have about schedule.

ISSN 2277 - 5080 | © 2016 Bonfring

Bonfring International Journal of Research in Communication Engineering, Vol. 6, Special Issue, November 2016 51

[1] P. Saini and P. Kaur, “Automatic speech recognition: A review”,

International journal of Engineering Trends & Technology, Pp.132-136,
2013.
[2] L. Deng and X. Li, “Machine learning paradigms for speech recognition:
An overview”, IEEE Transactions on Audio, Speech, and Language
Processing, Vol.21, No.5, Pp.1060-1089, 2013.

ABOUT THE AUTHOR

Chaitra Prakash Shet (30-08-1995) is an under graduating
student pursuing her final year degree in Electronics and
Communication department of SDMCET, Dharwad,
Karnataka. She is interested in VLSI, Control Systems and
HDL. (E-mail: [email protected])
Pooja Ravindra Nayak (17-05-1996) is an under graduating
Figure 4: Fields of AI student pursuing her final year degree in Electronics and
Communication department of SDMCET, Dharwad,
AI is (also known as intelligent system)mainly based upon Karnataka. She is interested in VLSI, Control Systems and
mathematics(particularly logic, combinatory, statistics, HDL. (E-mail: [email protected])
probability and optimization theory), psychology, linguistics,
neuroscience and philosophy [10].but it is primarily a branch
of computer science and it has borrowed a lot of concepts and
ideas from the above mentioned fields.

XI. ULTIMATE GOAL

The ultimate goal of the Artificial Intelligence is to help
user do his work simultaneously which in turn is going to
reduce the time consumption and also reduce human work to a
greater extent.

XII. CONCLUSION
This speaker recognition technology helps physically
challenged skilled persons. These people can do their works
by using this technology by consuming very less time. This
ASR technology is used in military weapons and in Research
centers. This technology is also used by CID officers in order
to trap the criminal activities.

REFERENCES
[1] J.L. Barrett and F.C. Keil, “Conceptualizing a non- natural space entity”,
Anthropomorphism in God concepts Cognitive Psychology, Vol.31,
No.3, Pp.219:247, 1996.
[2] Himanshu and S. Kaur, “Literature Survey on Automatic Speech
Recognition System”, Vol.4, No.7, 2014.
[3] H.H. Ammar, W. Abdelmoez and M.S. Hamdi, “Software engineering
using artificial intelligence techniques: Current state and open
problems”, In Proceedings of the First Taibah University International
Conference on Computing and Information Technology, Al-Madinah
Al-Munawwarah, Saudi Arabia, Pp.52, 2012.
[4] J.T. Chien and S. Furui, “Predictive hidden Markov model selection for
speech recognition”, IEEE Transactions on Speech and Audio
Processing, Vol.13, No.3, Pp.377-387, 2005.
[5] A.M. Anusuya and K.S. Katti, “Speech Recognition By Machine: A
Review”, International Journal of Computer Science and Information
Security, 2009.
[6] W. Dai and P. Wang, “Application of pattern recognition and artificial
neural network to load forecasting in electric power system”, In Third
International Conference on Natural Computation, Vol.1, Pp.381-385,
2007.
[7] A. Choudhary and R. Kshirsagar. “Process Speech Recognition System
using Artificial Intelligence Technique”, International Journal of Soft
Computing and Engineering, 2012.
[8] S. Rawat, P. Gupta and P. Kumar, “Digital life assistant using automated
speech recognition”, 2014 Innovative Applications of Computational
Intelligence on Power, Energy and Controls with their impact on
Humanity (CIPECH), Pp. 43-47, 2014.

ISSN 2277 - 5080 | © 2016 Bonfring