VOICE CONTROLLED ROBOT FOR SURVEILLANCE AND GAS LEAKAGE DETECTION

Transcription

1 VOICE CONTROLLED ROBOT FOR SURVEILLANCE AND GAS LEAKAGE DETECTION Mallikarjuna Gowda.C.P 1, Raju Hajare 2, Akhil Kumar 3,Manasa.R.E 4, Ramyashree.R 5, SmithaPatil 6 1,2 Associate professor, Department of TE, B M S Institute of Technology and Management, Bengaluru, India 3-6 VIII Semester students, Department of TE, B M S Institute of Technology and Management, Bengaluru, India Abstract- The objective of this project is to design a voice controlled robot for surveillance purpose and gas leakage detection. The system comprises of a user interface module and a Robot. The user interface includes a microphone attached to the PC. The software running on PC processes the signals to recognize the voice commands. The software is written using MATLAB. Microcontroller in the receiving side controls the motion of the Robot. The control commands and data are transmitted and received using ZigBee. Gas Sensor is present in the robot for detecting gas leakage. Index Terms- Voice controlled, Zigbee, User interface, Gas leakage detection, Matlab. I. INTRODUCTION Until recently, robots are making their way into social public services. The market for robots is expected to grow rapidly. instructions are given to the mobile platform via zigbee. The speech recognition software is speaker dependant. The special feature of the application is the ability of the software to train itself for the above voice commands for a particular user. It provides a very convenient method for the user to train the system. The camera in the robot is used for surveillance purpose and enables the operator to decide the direction of robot movement. Gas sensor in the robot detects gas leakage if any, and the GPS sends its exact location to the user interface. II. DESIGN The block diagram is as shown below. As the most natural and expressive means of communication, speech is a suitable choice for the human robot interaction. Applying speech in robot control would be very convenient. Speech control robot system has potential for application in somewhere voice communication plays a crucial role. The operator only needs to speak the command to make the robot to achieve the task. Voice Controlled Robot (VCR) is a mobile robot whose motions can be controlled by the user by giving specific voice commands. The speech recognition software running on a PC is capable of identifying the 5 voice commands Front, Back, Left, Right and Stop issued by a particular user. After processing the speech, the necessary motion Speech recognition system IJIRT INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 1656

2 MATLAB function filter. b) Frame Blocking: Each signal is now converted into a set of fixed length frames, with some number of samples in each frame is overlapping. If the frame length is L and each frame is shifted by M samples away from the adjacent frame, then nth frame can be denoted by, Word Capturing The signals coming from the microphone is processed only when you speak something. The program waits until the sample value exceeds some threshold value (which can be adjusted by the user). When the program is triggered by a significant sample, a number of following samples are captured to process. After that to determine the actual boundaries of the word spoken, edge detection is performed. Here the center of gravity of the energy distribution of the signal is calculated and then from that point intervals where the amplitude level lies below a threshold level are removed. Finally we can have a set of voice samples corresponding to a particular word free of silent periods. LPC Processing The steps we followed for the extraction of speech characteristics from captured samples using LPC is described in the following block diagram. Where n=1,2.n and i=1,2..l c) Windowing: Each individual frame is windowed to minimize the signal discontinuities at the borders of each frame. We used the Hamming Window for this purpose. The set of samples for each frame is multiplied by the time domain version of the Hamming window with size equal to the frame length. d) LPC Calculation: First step of calculating LPC parameters is to get the autocorrelation vector for each frame. If the order of the autocorrelation is P, then the autocorrelation vector, r can be given by, where m = 0,1,2,,P and x(i) s (i = 1,2,..,L) are sample values in the windowed frame. Then Hermitian Toeplitz matrix of r is computed as shown below: LPC Processing a) Pre-emphasis: This operation is necessary for removing DC and low frequency components of the incoming speech signal. It also makes the signal spectrum flatter. Pre-emphasis is done using a first order FIRfilter which can be described by the transfer function, Finally the LPC parameter matrix, a is calculated by matrix multiplication of inverse of R and r. Here we used a = 0.9. FIR filtering was applied to the signal in the time domain using the e) Cepstral Co-efficient Calculation: Cepstral co-efficients are Fourier transform representation of the log magnitude spectrum. Use IJIRT INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 1657

3 of Cepstral co-efficients makes the application more robust and reliable. background of this network and learning algorithm. Figure 4.4 shows the block diagram of the backpropagation training algorithm. Input Layer transfer function Hidden Layers & Output Layers transfer function As low order LPC parameters are too sensitive to the spectral slope and low order parameters are sensitive to noise, Cepstral co-efficients are weighted by a tapered window. So weighted set of Cepstral coefficients, Cm` are found by, For each frame belong to the captured word, this sequence of operations are performed and finally we end up with a number of sets of Cepstral coefficients. Next values for all the frames are averaged to get a single set of Cepstral co-efficients for that spoken word. Artificial Neural Network: To recognize the speech mobile robot should have intelligence. Hence artificial neural network is used to make intelligence through the learning process. In here we used Multi Layer Feedforward Network and error back propagation learning algorithm to train the network. Figure is shown the network model. Network Model This network consists of two Hidden layers with the Input & Output layers. Liner transfer function used to the input layer and sigmoid transfer function applied to hidden layer as well as output layer. Following section describe the mathematical Backpropagation training algorithm. This neural network builds on MATLAB computing environment and it process data in real time and the program give the outputs according to recognized voice command. First of all calculated set of Cepstral co-efficients for that spoken word is fed to input layer of ANN and calculate the output. During the training process we input the set of voice command samples with the relevant output network to be produced. For each iteration network calculate the actual network output and compare with the specified out and error is backpropagate to minimized the error in next iteration. Tuning Process The above Neural network we used is capable of identifying two voice commands. In order to identify more commands we used three separate networks. When using this kind of process we need to set output threshold levels according to the way that it had trained. To do this we give known voice patterns to the already obtained neural network and set the threshold levels accordingly. Graphical User Interface The main objective of the GUI is to provide a user friendly Interface which integrates all the required functionalities. That is it should be able to facilitate all phases of voice recognition algorithm including IJIRT INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 1658

4 acquiring voice samples, then train the neural network and finally running voice recognition process in real time. There are several important areas within the GUI. They are Mode Selection Panel (Used to select required Mode that user want to run) o Acquisition Mode Obtain voice samples separately for each command and then calculate Cepstral co-efficients of each voice sample and store them o Training Mode Use cepstral coefficients to build the Neural Network. So Neural Network coefficients are obtained. o Tuning Mode Use known voice samples to calculate threshold levels needed. o Running Mode - Each voice input is sampled real time and use the neural network to obtain the output and then use threshold levels obtained in the Tuning mode to recognize the voice input Options Panels (Acquisition, Tuning and Running) o Used to set options needed at each Mode Recognition Indications o This is enabled at Running Mode and used to indicate the status of the latest voice input. Graph Window o Used to display acquired sound pattern Emergency Stop o Used to issue a STOP command if needed Hardware construction The mobile robot platform consists of three wheels and backward two wheels are coupled to the two geared DC motors. The rest wheel can free rotate with its own axis. The robot is capable of moving forward, backward, left and right and onboard motor driver circuit with L298 IC is used to control the motion with the aid of PWM (Pulse Width Modulation) switching signals. Most popular low cost 8051 microcontroller is the main controller within the mobile robot which is communicates with the remote host computer through the wireless zigbee module. Zigbee is configured using TMFT(Tarang Multi Function Tool) software. Using this software we can set the source and destination address of zigbee. Remote host is acting as master and microcontroller on robot acting as slave to transmit asynchronous serial data with RS232 data format between these two devices. Hence according to the received command from the host, microcontroller generates relevant PWM switching signals according to controlling algorithm. This microcontroller has a special capability to generate HPWM (Hardware PWM) which helps to generate the two PWM signal simultaneously without affecting execution of other part of program. To establish the communication link between the mobile robot with the remote host computer, Tarang f4 zigbee module is used which is operating in the 2.4 MHz frequency band. The power supply for the mobile robot is big issue. Because DC motors consume huge power to overcome friction during the operation. Hence Sealed lead Acid rechargeable 12V & 1 ampere battery is used to power the mobile robot. The camera in the robot is used for surveillance purpose and enables the operator to decide the direction of robot movement. The power supply for the camera is given using 9V battery. It is a wireless AV camera. Zebronics TV tuner is used to display the received AV on the laptop. The radio AV receiver receiver basically refers to an amplifier that has a built-in radio tuner. With A/V receivers the basic functionality is to receive an audio signal, amplify the audio signal, and allow pass-through of the corresponding video signal to a display device such as a projector or a television. Mq-6 is a simple-to-use liquefied petroleum gas (LPG) sensor, suitable for sensing LPG (composed of mostly propane and butane) concentrations in the air. The MQ-6 can detect gas concentrations anywhere from 200 to 10000ppm. This sensor has a high sensitivity and fast response time. The sensor s output is an analog resistance. The drive circuit is very simple; all you need to do is power the heater coil with 5V, add a load resistance, and connect the output to an ADC. This Gas sensor in the robot detects gas leakage if any, and the GPS sends its exact location to the user interface. The latitude and longitude are displayed on the terminal window of laptop. The circuit diagram of robot is shown in Figure 4.5. IJIRT INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 1659

5 Two 16 bit timers/counters.32 bit bidirectional I/O lines can be either used as 4 8bit ports or 32 individually addressable I/O lines. Two 16 bit timers/counters. 16 bit address bus multiplexed with p0 and p2 and 8 bit data bus multiplexed with p0. circuit Diagram III. SYSTEM OVERVIEW a) Zigbee wireless technology c) GPS(Global Positioning System) L80 is an ultra slim module with embedded mm patch antenna. It is a SMD type module, with compact profile, ultra low power consumption and fast positioning capability. L80 can calculate and predict orbits automatically using the ephemeris data (up to 3 days) stored in internal flash memory, so L80 can fix position quickly even at indoor signal levels with low power consumption. A type of low cost, low power wireless technology which is used for the different purposes at ultra low power is known as ZigBee Technology. It was not proposed or designed for the high speed data transfer applications. But it was designed for working on excellent long battery timings at low cost and also at ultra low power consumption. b) 8051 microcontroller Features: On chip clock oscillator. IJIRT INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 1660

6 IV. RESULTS In this system the experimental results mainly focus on the usability and effectiveness of the system. A. Usability Proof The voice commands given by a single user are processed using MATLAB software and are sent to the robot via zigbee. Microcontrollers generate the corresponding control code and outputs them to the motor driver and the Robot moves accordingly. The required parameters (location of gas leakage, camera output) are collected and send back to the user and is displayed on the PC. ZigBee provides high noise immunity due to Spread Spectrum modulation. B. Effectiveness Proof The effectiveness test is done by focussing on the correctness of the final result. We have tested under two different environments which are: 1) a quiet room 2) a noisy room. The robot performs more accurately in quiet room than noisy room. V. CONCLUSION Based on the experiments results, we can conclude that the Robot moves as per the user voice commands in all different environmental conditions and is able to send the desired data to the PC. The only disadvantage of this system is that the Robot may misinterpret words having same frequencies and operate in the wrong way. Words have to be said in such a way that they are of different frequencies. REFERENCES [1] Voice Controlled Data Acquisition Car Based on Zigbee Technology, Dinu Mathew, ME, Electronics and TeleCommunication, Sathyabama University, Chennai, India. IOSR Journal of Electronics and Communication Engineering (IOSRJECE) ISSN : Volume 2, Issue 3 (July-Aug 2012), PP [2] Voice Communication Based on ZigBee Wireless Sensor Networks, LIU Meiqin, WANG Yuxuan, FAN Zhen, ZHANG Senlin. Proceedings of the 33rd Chinese Control Conference, July 28-30, 2014, Nanjing, China. [3] Xiaoling Lv, Minglu Zhang and Hui Li, Robot Control based on Voice Command, The Proceeding of the IEEE International Conference on Automation and Logistics, China [4] Chomtip Pornpanomchai, Thammarat Saengsopee and Teravit Wongseree, Robot Arm Control by Using Thai Voice Commands, The 1st Northeastern Computer Science and Engineering Conference, Thailand, [5] Parichart Leechor, Chomtip Pornpanomchai, Operation of a Radio-Controlled Car by Voice Commands, 2nd International Conference on Mechanical and Electronics Engineering, Thailand, [6] Zigbee/IEEE Networking Examples, Zigbee Wireless Network and Transceivers, IJIRT INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 1661