Dr.

Vasantraodada Patil Shetkari Shikshan Mandal’s

PadmabhooshanVasantraodada Patil Institute of Technology

Department of ELECTRONICS AND TELECOMMUNICATION.

SEMINAR REPORT ON

“SPEECH RECOGNITION USING ARTIFICIAL NEURAL

NETWORK.”

Under the guidance of

Prof.D.B.Kadam sir

Submitted by

Roll no Student Name

1357 Tanvi Balkrishna Ghatage

Academic Year :– 2023-2024

 Padmabhooshan Vasantraodada Patil Institute of Technology,
Budhgaon
Department of Electronics And Telecommunication Engineering.

CERTIFICATE
This is to certify that he have satisfactorily submitted a report on

“SPEECH RECOGNITION USING ARTIFICIAL NEURAL

NETWORK.”

Submitted by:-

Name : Tanvi Balkrishna Ghatage

Roll no : 57

Batch : S3

Class : SY. Btech

Incharge H.O.D
Prof.D.B.Kadam Dr.M.S.Chavan
 ACKNOWLEDGEMENT

During the selection of topic entitled as "Speech Recognition Using Neural

Network ", The help I received from our professors, family, and friends is
invaluable and I’m forever indebted to them. I would first like to express my
gratitude to our Honorable Principal Mr. B.S.Patil, Our HOD Dr.M.S.Chavan and
our Seminar Guide Prof.D.B.Kadam for their immense support, suggestion,
encouragement and interest in my seminar work. Without their invaluable
suggestions my seminar selection would be incomplete.

Last but not least, I would like to thank my friends, parents for their belief and
patience in my endeavour.

Tanvi Balkrishna Ghatage

Table of Contents

Abstract…................................................................................................................................

List of Figures…........................................................................................................................

Introduction….........................................................................................1
1.1 Definition….............................................................................................
1.2 Speech Recognition ……………………………………………

Literature Survey
2.1 Speech Recognition Process……………………………………….………
2.2 Structure of Standard Speech Recognition System….……………
2.3 Types of Speech Recognition System…………………………….…

System Analysis
3.1 Speech Recognition Algorithms……………………………………..…….
3.2 Dynamic Time Warping………………….……….…………….…
3.3 Hidden Markov Model……………………………………………
3.4 Neural Network…………………………………………………

Discussion
4.1 Speech Recognition Software………………………………………
4.2 Advantages & Disadvantages……………………………………
Advantages.……………………………………………………………...
Disadvantages……………………………………………………………
4.3Applications………………………………………………………………

Conclusion & Future Scope

5.1 Conclusion……………………………………………
5.2 Future Scope……………………………………
5.3 Acknowledgement…
 List of Figures

1. Hidden Markov Model

2. Speech Recognition

3. Speech Recognition System

4. Out of Vocabulary

5. Dynamic time warping

6. Automatic Speech Recognition

7. Operating System

8. Large Vocabulary Continuous Speech Recognition

9. Intelligent Rival Imitator of SIRI

 ABSTRACT:

Language is man's most important means of communication and speech its primary
medium. Spoken interaction both between human interlocutors and between humans
and machines is in escapable embedded in the laws and conditions of
Communication, which comprise the encoding and decoding of meaning as well as
the mere transmission of messages over an acoustical channel. Here we deal with this
interaction between the man and machine through synthesis and recognition
applications. Speech recognition, involves capturing and digitizing the sound waves,
converting them
to basic language units or phonemes, constructing words from phonemes, and contex
tually analyzingthe words to ensure correct spelling for words that sound
alike. Speech Recognition is the ability of a computer to recognize general, naturally
flowing utterances from a wide variety of users. It recognizes the caller's answers to
move along the flow of the call. Emphasis is given on the modelling of speech units
and grammar on the basis of Hidden Markov Model& Neural Networks. Speech
Recognition allows you to provide input to an application with your voice. The

applications and limitations on this subject enlighten the impact of

speech processing in our modern technical field. While there is still much room for
improvement, current speech recognition systems have remarkable performance. We
are only humans, but as we develop this technology and build remarkable changes
we attain certain achievements. Rather than asking what is still deficient, we ask
instead what should be done to make it efficient.
INTRODUCTION

Speech recognition allows you to provide input to a system with your voice. Just like
clicking with your mouse, typing on your keyboard, or pressing a key on the phone
keypad provides input to an application, speech recognition allows you to provide
input by talking. In the desktop world, you need a microphone to be able to do this.

1.2 SPEECH RECOGNITION

Speech recognition (or sometimes referred to as Automatic Speech Recognition) is

the process by which a computer (or other type of machine) identifies spoken words.
Basically, it means talking to a computer & having it correctly understand what you
are saying. By “understand” we mean, the application to react appropriately or to
convert the input speech to another medium of conversation which is further
perceivable by another application that can process it properly & provide the user
the required result. The days when you had to keep staring at the computer screen
and frantically hit the key or click the mouse for the computer to respond to your
commands may soon be a things of past. Today we can stretch out and relax and tell

your computer to do your bidding. This has been made possible by the ASR

(Automatic Speech Recognition) technology. Speech recognition is an alternative to

traditional methods of interacting with a computer, such as textual input through a
keyboard. An effective system can replace, or reduce the reliability on, standard
keyboard and mouse input.
This can especially assist the following:

 People who have little keyboard skills or experience, who are slow typists, or
do not have the time or resources to develop keyboard skills.
 Dyslexic people or other who have problems with character or words uses and
manipulation in a textual form.
 1.1 What is a Smart Sensor System?

1. Network Capable Application Processor (NCAP) where control and data correction takes
place
2. Transducer Interface Module (TIM) (one or more) containing the transducer and
data acquisition

NCAP (Network Capable Application Processor)

 Communications
 Interface Control
 Message Routing
 TIM Discovery and Control
 Data CorrectionInterpretation of TEDS Data

 Message Encoding and Decoding

TIM (Transducer Interface Module)

Analog Signal Conditioning
Triggering
Analog to Digital Conversion
Command Processing
TEDS Storage
Data Transfer
Communications
Figure 1.2(a): A Smart Sensor System
 Figure 1.2(b): A Smart Sensor System

1.2 IEEE 1451 – A Standard for Smart Sensors

What is IEEE 1451?

An industry-wide open standard for Intelligent Sensors (Smart Sensors)

• Develop network-independent and vendor-independent transducer interfaces
• Support a general model for transducer data, control, timing, configuration and
calibration Specifies physical and functional interfaces between sensors/actuators and
instruments/ microprocessors/networks
Specifies analog, digital and wireless interfaces to provide ease in connecting sensors and
actuators either by wireline or wireless methods
Provides a self-describing sensor via the Transducer Electronic Data Sheet (TEDS) that contains
manufacture-related data
Allows sensors to be installed, upgraded, replaced and/or moved with minimum effort
Eliminates manual entering of data and system configuration steps which is error prone.
Chapter 2
General Architecture of smart sensor

One can easily propose a general architecture of smart sensor from its definition, functions.
From the definition of smart sensor it seems that it is similar to a data acquisition system, the
only difference being the presence of complete system on a single silicon chip. In addition to this
it has on–chip offset and temperature compensation. A general architecture of smart sensor
consists of following important components:
 Sensing element/transduction element,
 Amplifier,
 Sample and hold,
 Analog multiplexer,
 Analog to digital converter (ADC),
 Offset and temperature compensation,
 Digital to analog converter (DAC),
 Memory,
 Serial communication and
 Processor
The generalized architecture of smart sensor is shown below:
 Figure 2.1: Internal Architecture of Smart Sensor

2.1 Description of Smart Sensor Architecture

Architecture of smart sensor is shown. In the architecture shown A1, A2…An and S/H1,
S/H2…S/Hn are the amplifiers and sample and hold circuit corresponding to different sensing
element respectively. So as to get a digital form of an analog signal the analog signal is
periodically sampled (its instantaneous value is acquired by circuit), and that constant value is
held and is converted into a digital words. Any type of ADC must contain or proceeded by, a
circuit that holds the voltage at the input to the ADC converter constant during the entire
conversion time. Conversion times vary widely, from nanoseconds (for flash ADCs) to
microseconds (successive approximation ADC) to hundreds of microseconds (for dual slope
integrator ADCs). ADC starts conversion when it receives start of conversion signal (SOC)
from the processor and after conversion is over it gives end of conversion signal to the
processor.
Outputs of all the sample and hold circuits are multiplexed together so that we can use a single
ADC, which will reduce the cost of the chip. Offset compensation and correction comprises of
an ADC for measuring a reference voltage and other for the zero. Dedicating two channels of
the multiplexer and using only one ADC for whole system can avoid the addition of ADC for
this.
This is helpful in offset correction and zero compensation of gain due to temperature drifts of
acquisition chain. In addition to this smart sensor also include internal memory so that we can
store the data and program required.
Chapter 3
Block Level Design Considerations for Smart Sensor

Design choice of smart sensor depends on the specific application for which the sensor is
required and also related to specific industry. Normally a smart sensor will utilize inputs form
one or more sensor elements either to generate an output signal or to generate a correction
signals which are applied to the primary output. This includes design of circuitry to take output
of raw sensor elements and generate compensated and linearized sensor output.

3.1 Functions within electronics:

The smart sensor contains some or all of the following functions
 Sensor Excitation:
Many a times it is required to alter the sensor excitation over the operating range of a
sensor. An example of this is a silicon wheatstone bridge, where the drive voltage is
increased with increasing temperature. This is done to compensate for the reduction in
sensitivity of the piezoresistors with increase in temperature. A drive stage with
temperature dependence can be used which is control by a microprocessor. This will also
reduce the calibration time.
 Analogue Input:
Multiplexing of inputs can be done to avoid duplication of circuit. In multiplexing inputs
of same type and range are switched to a common front end. The outputs of sensors are
normalized before they are switched and a variable gain stage is included after the
multiplexer. This allows the sensitivity variations between the different sensors to be
accounted for by a common front-end .In addition to this an offset adjustment is also
included in the common front end. The variable gain stage also offers an additional
advantage where the input signals are to be sampled by
 analog to digital converter (ADC) with fixed reference points. Under such situation gain
can be increased at the lower end to increase the sensitivity.

 Data Conversion:
In case of smart sensor most of the signal processing is done in digital form. This is
possible only when we
have an ADC along with an anti-aliasing filter. This is because most of the sensor output
is in the analog
form. Choice of ADC depends on the resolution, bandwidth and complexity of anti-
aliasing filter.
 Digital data bus interface:
The controller embedded in the smart sensor supports communications by digital data
bus. The advantages
of this are:
Wiring is reduced considerably
Automatic calibration at production can be simplified.
 Monitoring and diagnostic functions:
In many applications self-test is required. This self-test includes connectivity checking
and long-term offset
correction.
 Control processor:
To provide greater flexibility and reduced complexity, a control processor can be used.
Control processor
can do digital filtering. Another important point is software development. Processor must
allow writing
codes in higher language as it reduces the development time.

3.2 Level of integration:

Though it is possible to integrate smart sensor on a single piece of silicon it is unattractive due to
cost and performance. Analog processing, digital logic and non-volatile memory (NVRAM), can
all be done on same piece of silicon. But compromise must be made that limit the performance of
at least one of these functions.

Chapter 4
DESIGN

There are three important aspects in our design { data acquisition, data transfer and data
processing}. Major priority is to minimize the total cost of the system while at the same time
delivering the amount of accuracy and reliability provided by the proprietary solutions. The
system will also be generic enough to be adapted to other areas of environmental
measurements. The system synthesizes Field Programmable Gate Arrays and sophisticated
IEEE 802.11 wireless networking infrastructure adapted to environmental sensing.

4.1 OVERIEW

We divide the system into three important components each handing a specific task.
This modular design accounts for the possibility of incorporating new advancements
in each component without disturbing the whole system. Each component will have well
defined interfaces. The components are the Data Acquisition Unit (DAU), the Data Transfer
Unit (DTU) and, the Data Processing Unit (DPU). As seen earlier, The DAU is responsible
for collecting the data from the sensor and responds to data requests from the DTU. The
DTU will transport the data from the DAU to the DPU reliably. Finally, the DPU, after
gathering the data from sensors, will analyze and store the data for end use.

4.2 Data Acquisition Unit

The DAU will consist of a sensor that is connected to a micro-controller implemented on a

FPGA through an Analog to Digital Converter (ADC). The sensor will measure the pressure,
which isrelated to the amount of water above it, and sends the analog signal to the ADC.
 The ADC will convert the analog signal to a digital value and sends the data to a
micro-controller when requested. which we have developed. During the course of
further improvement, we will need to consider the extent to which we will include the
computation of temperature compensation in the DAU. The power for the devices
can be provided by using regular A/C power supply with adapter, solar panels or
long-life batteries.

Sensor ------- ADC ------ FPGA---------Antenna

4.3 Data Transfer Unit

The data transfer unit is responsible for carrying the data from the DAU to the DPU.
Data transfer will be accomplished IEEE 802.11 wireless network protocols. Cellular
technology, which is an alternative solution, enables the connectivity between remote
locations and the Internet. Cellular phone service currently has a limitation of avail-
ability and stability in remote locations across Nebraska. A high gain IEEE 802.11
network is potentially a better choice because it costs less and the user has more
control. The IEEE 802.11 network will utilize TCP/IP network protocol thereby
providing the ability to control the DAU to a remote operator. The use of intelli-
gent routing protocols will enable sensors to form networks and thereby routing the
information in the absence of direct line of sight to the base station. Our goal is to
transfer the data in real-time (recall that once per day is considered to be real time)
to the Data Processing Unit (DPU). The challenges that have to be addressed during
DTU development are the stability of connectivity and data security. Data security
can be provided by IEEE 802.11 Wired Equivalent Privacy (WEP)

4.4 Data Processing Unit

The data processing unit is where the data can be manipulated by analysts and/or
made available to the public. New archiving and analysis techniques need to be
developed to handle the accumulation of potentially large amount of data every day.
Most components of the DPU will consist of software that will allow the user to
access the data graphically (see Fig. 3.3) or in a tabular format. In future, maps
may be considered for data display. A key characteristic of any software that will
be developed is that it be °exible and user-friendly. The development of the DPU
targets these two goals. For the research to be complete, correct and e®ective, collaboration and
cooperation among the experts in computer technology and water resource management( in case
of water management application) isrequired. The design of such a system needs to be exible
enough so that it can be scaled to a variety of applications and for it to provide stable
connectivity.

Figure 4.1: Data Processing Un

 REFERENCES

[1] Dayashankar Dubey,”Smart Sensors”, ElectronicSystemsGroup,IIT Bombay,November2002.

[2] Srikanth Anumalla,” GROUNDWATER MONITORING USING SMART SENSORS”,

Lincoln, Nebraska, August, 2004.
[3] Wilfried Elmenreich,” Smart Transducers - Principles, Communications, and Configuration”.

[4] W. H. Ko and C. D. Fung. VLSI and intelligent transducers.

Sensors and Actuators, (2):239–250, 1982.

[5] Harish Ramamurthy, B. S. Prabhu and Rajit Gadh,” Smart Sensor Platform for Industrial
Monitoring and Control”, Wireless Internet for the Mobile Enterprise Consortium
University of California, Los Angeles.

[6] Alex Karolys,” A Smart Network Sensor System for Distributed, Synchronous Data
Acquisition”, VIP Sensors, Inc. – 27128A Paseo Espada, St. 1523 - San Juan Capistrano, CA
92675

[7] IEEE 1451.2 - A SMART TRANSDUCER INTERFACE for SENSORS and ACTUATORS
- Transducer to Microprocessor Communication Protocols and Transducer Electronic Data Sheet
(TEDS) Formats.

[8] Perry Sink, “A Comprehensive Guide to Industrial Networks, Part 1: Why Use an Embedded
Network or Fieldbus, and What Are the Most Popular Standards?” Sensors Magazine, June 2001
http://archives.sensorsmag.com/articles/0601/28/index.htm
PVPIT Budhgaon, Department of E&TC Engineering Page 28
PVPIT Budhgaon, Department of E&TC Engineering Page 29
PVPIT Budhgaon, Department of E&TC Engineering Page 30