HANDICAPPED VOICE RECOGNITION CONTROL SYSTEM BASED ON HMM ALGORITHM

Transcription

1 International Journal of Mechanical Engineering and Technology (IJMET) Volume 9, Issue 11, November 2018, pp , Article ID: IJMET_09_11_110 Available online at ISSN Print: andISSN Online: IAEME Publication Scopus Indexed HANDICAPPED VOICE RECOGNITION CONTROL SYSTEM BASED ON HMM ALGORITHM Bashra Kadhim Oleiwi, Farah F. Alkhalid and Abdul Muhsin M. Control and Systems Engineering Dept. /University of Technology- Iraq ABSTRACT This paper addresses the voice recognition problem of handicapped people based on HMM algorithm and provides a system that let disabled people and elders to manage part of their life independently, like turn on and off some electric devices in home by voice commands. Arduino Uno is used as microcontroller which senses a voice command via microphone connected in voice recognition module that compatible with Arduino Uno, and finally Relay to make interface between 5volts of Arduino and 220 volts of Iraq Alternating Current (AC). To verify the credibility of the proposed system, the testing results of the system have been provided, which meets its desired goal. Keywords: Voice recognition, Arduino, HMM algorithm. Cite this Article: Bashra Kadhim Oleiwi, Farah F. Alkhalid and Abdul Muhsin M, Handicapped Voice Recognition Control System Based On Hmm Algorithm, International Journal of Mechanical Engineering and Technology, 9(11), 2018, pp INTRODUCTION Since the early manual wheelchairs to modern electric-powered counterparts, developments have focused on improving autonomy in mobility. Many research projects introduced different solutions in this field, such as [1] proposed a control system that used by elderly, handicapped patients, and people with disabilities to control all appliances and devices from anywhere in the house remotely using XBee sensor, this study did not take into account the disabled who would lose his arm control gradually. While [2] provided system which controls wheelchair using voice commands, this work was very useful to the disabled has weakness in upper limb. Also, as can be observed in [3] a humanoid robot which is capable of detecting and recognizing different faces or objects and can be controlled through voice. In [4] proposed a study in "Controlling electronic devices remotely by voice and Brain Waves While, in [5] introduced a study in the voice-commands that are recognized by dedicating hardware module and recognizing data which sent wirelessly through a radio-frequency (RF) transmitter attached to a Arduino microcontroller board. In [6] proposed system used voice controlling to enable centralize controlling over distant household appliances using Kinect Sensor editor@iaeme.com

2 Handicapped Voice Recognition Control System Based On Hmm Algorithm This study provides a way to let disabled people and elders managing part in their life independently by turning on and off some electric devices at home based on voice commands, such as shutting down the air-condition whenever it is cold, or may turning on TV whenever, when it is boring, just by voice command. Arduino Uno is used as microcontroller which senses a voice command via microphone connected in voice recognition module which is compatible with Arduino Uno, and finally relay to make interface between 5volts of Arduino and 220 volts of Iraq AC. In this study, the proposed system explained, and then the system model and implementation presented, finally results and conclusions discussed. 2. PROPOSED SYSTEM DESCRIPTION In order to explain the current proposed system, it is divided into two stages. The first stage is related to the preparation and connection of hardware components and settings of proposed, which is called hardware stage, while the second stage describes the operations of proposed system, which is called software stage. A. Hardware Stage This stage consists of hardware components, which are related to the connection of circuit according to the present description of the problem, as can be seen in figure (1) the block diagram of proposed system which consists of Arduino Uno, voice recognition module, Relay and wires. Where the Arduino Uno operates or shuts down the electrical devices by giving voice commands stored in a voice recognition module via relay that makes compatible interface between 5 volts (Arduino Uno) and 220 volts AC. Thus, a paralyzed and disabled person can turn on and off the electrical equipment s depending on himself. Figure (1): Block Diagram of the proposed System. The mechanism of this system is the handicapped person who cannot serve himself because of his limited movement, can only give an order to let devices turn on or off, this proposed control system can save several voice commands in the voice recognition module to be used later to control several operations editor@iaeme.com

3 Bashra Kadhim Oleiwi, Farah F. Alkhalid and Abdul Muhsin M 1. Microcontroller Arduino Uno microcontroller has used as the processing core of this control system. It is an electronic development board, based on hardware, and software, easy to use and learn. It can be used by both professionals and beginners, and it uses everything you can think of. It can be observed in table (1) the main characteristics of this microcontroller [7]: Component Table (1): Arduino Uno characteristics [7]. Characteristic Microcontroller ATmega 328 Working voltage Input voltage limits I / O outlet Current voltage in the 3.3V pin Memory Size Speed 5 volts 6-20 volts, preferably 7-12 volts. 40 ma 50 ma 32 KB 16 MHz 2. Voice Recognition Unit Voice recognition V3 is an easy-to-use control panel that is used to identify voice/ speech to control different projects. The V3 unit has the capacity to store up to 80 voice commands each with 1500 milliseconds (approximately one or two words per command). They are stored in one large collection such as the library. But during its work it cannot distinguish between the 80 sounds at the same time, so seven voice commands are imported / recalled into the library to the recognizer, this means that maximum voice commands are seven effective commands at the same time [8], Table (2) lists the characteristics of voice recognition unit. Table (2): Voice Recognition Module characteristics [8]. Component Voltage Current Digital Interface Analog Interface Size Characteristic V <40mA 5V TTL level UART interface 3.5mm mono-channel microphone connector + microphone pin interface 30mm x 47.5mm 3. Relay The Relay is an "OFF - ON" cutter circuit, but it closes on or opens off not by hand as in the regular cutouts, but by applying a suitable electric voltage on the relay (either is 6 volts or 9 volts, or 12 volts or 220 volts) [9]. The Relay uses to convert from 5 volts (the output voltage of Arduino controller) to 220 volts (required voltage to operate electrical devices according to national AC electricity in Iraq), the table (3) shows the characteristics of the relay: editor@iaeme.com

4 Handicapped Voice Recognition Control System Based On Hmm Algorithm Component Table (3): Relay characteristics [9]. Characteristic Supply voltage +5V Supply current Current on pin IN Rated load Operating temperature Operate time max. Release time max. Insulation resistance 72mA typ. (75mA max.) 14mA typ. 7A 250VAC -30 C / +70 C 10ms Max. 5ms Max. 100Mohm Min. B. Software Stage Using Arduino IDE (Integrated Development Environment) to program the algorithm for saving and then recognizing the voice order based on Hidden Markov Model (HMM algorithm). 1. HMM Algorithm The main goal of speech recognition system is to convert an input speech signal into its corresponding wave form or textual form using some algorithm as a computer program [10].HMM is a probabilistic pattern matching technique in which the observations are considered to be the output of stochastic process and consists of an underlying Markov chain [11]. HMM provides a simple and effective framework for modelling time-varying spectral vector sequences [13]. First of all, the waveform of continuous speech is partitioned into many frames each frame has the same length (approximately 25ms). For each spoken word, let O = o1 o2 oτ be a vector of sequence parameters, where to is spotted at time t {1, τ}. The D is a dictionary words, word wi D, the recognition equation is summarized by: =arg ( ) (1) Where is the recognized word, while P is the probability measurement finally a word sequence is W = w1 wk. Bayes Rule allows transforming (1) in a convenient calculable equation: ( )= ( )() () Where the acoustic model is denoted by P (O W) and the language model is P (W); P (O) can be neglected. Therefore, for any set of prior probability P (W), the likelihood P (O wi) is used to make the most probability of spoken word. It is very important to remember that the mixture of the language model probability and the caustic model probability is weighted. a problem appeared is that HMM acoustic models usually underestimate the acoustic probability giving to the language model little weight. A solution for this problem is to increase a weight to grow the language model probability. Hence, the LM is frequently scaled by s, where s is an empirically determined constant, called language model scale factor (LMSF). LMSF is determined empirically to make optimize performance in the recognition. This weighting has a side effect as a penalty for inserting new words. Therefore, scaling factor pis added, that the price of word insertions called word insertion penalty (WIP) and can be calculated empirically. Thus (1) becomes: (2) editor@iaeme.com

5 Bashra Kadhim Oleiwi, Farah F. Alkhalid and Abdul Muhsin M =arg ( )() W (3) In the log domain, the final likelihood is calculated as: = + P (W) +p (4) Where W is the length of the word sequence W, s is the language model scale factor (LMSF) (approximately in the range of 5 to 15) and p the word insertion penalty (WIP) (approximately in the range of 0 to -20). 2. Mechanism of the Proposed System The mechanism of the proposed system is illustrated in flowchart in figure (2). Thus, the proposed control system will be operated and listened, if there is a new voice command. Whenever a specific voice command can be detected, so the system would simply operate or shutdown the device, and then the disabled person will feel that there are supernatural powers which respond to his orders instantaneously. If the voice command does not recognize by the system, so the system should listen to a new command and so on. Figure (2): Flowchart of the proposed control system. 3. Proposed System Implementation In order to verify the credibility and effectiveness of the proposed system, a implementtion was carried out in this section. The main three parts of this control system (Arduino Uno, voice recognition module and Relay) should be connected via wires as shown in figure (1), where the connection between Arduino Uno and the voice recognition module is represented by the pins which have fixed in table (4) editor@iaeme.com

6 Handicapped Voice Recognition Control System Based On Hmm Algorithm Table (4): Connection between Arduino Uno and Voice Recognition Module. Arduino Uno Voice Recognition 5v 5v 2 Tx 3 Rx GND GND While the pins connection between Arduino Uno and Relay are listed in table (5). Table (5): Connection between Arduino Uno and Relay. Arduino Uno Relay Both ends of the engine Both ends of COM Positive battery terminal End NO Negative battery terminal End NC Vcc Vcc Gnd GND Port 12 IN1 Port13 IN2 The testing result of the proposed system during running is demonstrated in figure (3). Figure (3): Implementation of the proposed system. The system was tested by saving order to operate device using woman voice with 35 years old. Thus, the wave form shows her order in figure (4) entitled Original voice command1, editor@iaeme.com

7 Bashra Kadhim Oleiwi, Farah F. Alkhalid and Abdul Muhsin M then other attempt to save voice order which done by man with old 70 years old, boy with 6 years old and boy with 3 years old as shown in figure (4) entitled Unmatched command respectively, the attempt was failed to recognize by the proposed system, only the owner of the saved wave form can control the system. Moreover, the proposed system tested experimentally with noisy voice command of the owner as can be noticed in figure (4) entitled noisy voice command, the response of the proposes system depends on the degree of the added noise to the original voice signal. The voice print as same as finger print, it is unique and the HMM algorithm interests in security biometric. Figure (4): Different waveform of voice. The difficulty of knowing the speech lies in the diversity of the form of the verbal sign that corresponds to the word itself. These changes can be classified as follows: 1. Changes related to the speaker: Changes in the voice can occur according to the physiological and psychological changes of the person (whether he is sick or true, sad or joyful or angry), according to the rate of his speech (fast or slow), and according to the quality of his voice or speak in an indirect way) editor@iaeme.com

8 Handicapped Voice Recognition Control System Based On Hmm Algorithm 2. Changes between speakers: follow the social background language to them, such as the tone, dimensions and form of their vocal acoustic device vocal traction. 3. Phonemes: the smallest units of sound in which words can be composed, and this investigation depends heavily on the context in which it is uttered. For example, the word b / from the word "door" is different from the word "sat"; it is in the first / b / and in the second / p /. 4. Speaker circumference: This is the result of undesirable voice signals (other speakers, noise, closing of door...), or the variety of mics used and where they are placed. This model requires a pre-recognition training process that involves drawing reference attributes for the individuals we want to know. In order to identify a single object, we balance its attributes with the reference attributes of all the reference terms to search for the closest item. In order to obtain an optimal knowledge, it is possible to benefit from the vocal and lexical limitations and linguistic models of the language used. 4. CONCLUSION The testing results of the design provide clear evidence of efficiently and effectively meeting the desired objective. The proposed system has implemented and observed that the voice command can sensed easy and fast, the device responded immediately, there is a very simple delay of less than one second, and the reason behind this delay is that there is a specific period for the purpose of completing one voice command, maximum 1500 milliseconds, If the stored instruction is less than this period, the system will certainly listen to the maximum period specified for each instruction, thus showing delay in milliseconds. In the other hand, the system proved reliable results with no errors, and can operate many different devices, reached to 80.The voice recognition model which was programmed utilizing HMM algorithm has sensed only saved voice tone, so the same voice order by different persons do not recognized because it saves the wave tone of voice order and not convert it to word. REFERENCES [1] G., Bilal and K., Khaled, Smart Home Automation System for Elderly, and Handicapped People using XBee, IEEE International Journal of Smart Home /ijsh , [2] N. J.Gor and A. Jeyakumar, Voice Controlled Motorized Wheelchair with Real Time Location Monitoring, IEEE International Conference on Energy, Communication, Data Analytics and Soft Computing (ICECDS-2017) /17, [3] U B. Sai, K. Sivanagamani, B. Satish, M. R. Rao, Voice controlled Humanoid Robot with artificial vision, IEEE International Conference on Trends in Electronics and Informatics ICEI 2017, /17, [4] D. Boucha, A. Amiri, D. Chogueur, Controlling Electronic Devices remotely by voice and Brain Waves,IEEE International Conference on Mathematics and information Technology, Adrar, Algeria December 4-5, 2017, /17, [5] K. V. Sai Vineeth, B. Vamshi and V. K. Mittal, Wireless Voice-Controlled Multi- Functional Secure ehome, /17, IEEE, [6] K. A. S. V. Rathnayake, W. H. K. P. Nanayakkara, and H. N. Gunasinghe, Voice Operated Home Automation System Based on Kinect Sensor, IEEE, International Conference on Information and Automation for Sustainability (ICIAfS), [7] About Arduino Uno (7, July 2018), available at: [8] [9] Voice Recognition manual (7, July 2018), available at: editor@iaeme.com

9 Bashra Kadhim Oleiwi, Farah F. Alkhalid and Abdul Muhsin M [10] [11] About Relay (7, July 2018), available at: [12] [13] S. Boruah and S. Basishtha, "A study on HMM based speech recognition system," IEEE International Conference on Computational Intelligence and Computing Research, Enathi, pp. 1-5, [14] W. H. Abdulla and N. K. Kasabov, The Concepts of Hidden Markov Model in Speech Recognition: Technical Report TR99/09, Department of Knowledge Engineering Lab Information Science Department University of Otago New Zealand, [15] M. Gales and S. Young, The Application of Hidden Markov Models in Speech Recognition, Foundations and Trends in Signal Processing, vol. 1, no. 3, pp , editor@iaeme.com