FORM 2
THE PATENT ACT 1970
(39 OF 1970)
AND
The patent rules, 2003
COMPLETE SPECIFICATION
(See section 10: rule 13)
1. TITLE OF INVENTION
Asthma Detector Using Fuzzy Neural based ASR system
2 APPLICANTS
Name Nationality Address
Ranjit Krishnarao Sawant Indian Department of Electronics and
Telecommunication, Government, Polytechnic , Hingoli
Plot No. 418 R.K.Nagar Society No. 6 Pachgaon, Kolhapur 416013, India.
Rajeshree Digambarrao Raut Indian Department of Electronics Design and
Technology
ShriRamdeobabaCollege of Engineering
and Management, Katol Road,
Gittikhadan, Nagpur - 440013
Ashok Anandrao Ghatol Indian Former Vice Chancellor, Dr. BATU,
Lonere ,
Saikrushna, Lane 2C,Prathmesh Park,
Baner-Balewadi road,
Baner, Pune, 411045

Technical field of invention:
This Invention relates to throat disease detection using Fuzzy Neural based Automatic Speech Recognition (ASR) system. The disease chosen for diagnosis is asthma. The tool is extremely user friendly and is an emergency assist to detect the asthma attack before the physician's arrival.
Background:
Throat diseases such as asthma, throat cancer, laryngopharyngeal reflux, phlegm, etc. are spreading a lot today. Disorder of the airways associated with increased airway hyper-responsiveness, recurrent episodes of wheezing, breathlessness, chest tightness, and coughing, particularly at night and early morning, such diseases can disturb the sound quality of the voice. The voice is generated by the mucus-covered muscular bands, which are the vibrating string that produces sound. When one has any kind of throat disease, these strings are filtered and shaped by resonating cavities of the throat, nose and mouth. Inflammation along the passageways from the nose down to the larynx, impair vocal quality. Some of the diseases also cause these strings to swell. If the strings are swollen then they don't vibrate, creating irrational voice.
Several methods for assessing speech pathologies have been introduced. The ASR task has various perspectives and face many challenges. The ASR system consists of two major blocks, viz. the feature extraction unit and classification technique. Popular feature extraction methods are, Linear Predictive Coding Coefficient (LPCC) and the Mel frequency Ceptstral Coefficient (MFCC). Five different classification methods can be used: template-based approaches, knowledge based approaches, artificial neural networks (ANNs), dynamic time wrapping (DTW) and hidden Markov models (HMMs). A Fuzzy Neural Network(FNN)-based system, which is widely employed in ASR systems, is the adaptive neuro fuzzy inference system (ANFIS). This system applies several fuzzy inference techniques for data classification. Several fuzzy rules for diagnostic respiratory infection are made in accordance with doctor's experience. The knowledge is adopted and extracted for building expert system. With the application of fuzzy-based expert system for respiratory infection diagnostics, the accuracy in diagnosis and analysis increases. Along these lines, our proposed model works using novel method to keep track of patient's pathology. It is easy to use, fast, non-invasive for patient and affordable for the clinician. It uses the parametric method like jitter and noise to evaluate the pathological voice. Another point on which the system relies is the Mel Frequency Cepstral Coefficient (MFCC) as feature extraction and ANFIS for feature mapping. The main aim is to evaluate the voice quality of the patients with symptoms of asthma and detect whether the patient is asthmatic or non-asthmatic.

Need of FNN-based Asthma Detector:
World health Organization(WHO)-Reports
■ More than 3 million people die every year from Respiratory Tract Infection-RTI (6% of the worldwide death)
■ 235 million people suffer from asthma; 251 million people suffer from Chronic Obstructive Pulmonary Diseases (COPD).
WHO predicts that COPD will become third leading cause of death worldwide. As per the news reports , published on July 01, 2015 by the Hindu, India tops world in Lung disease deaths.
Inspite of the above facts and figures RTI is under diagnosed and under treated.
■ Auscultation is a procedure used by physician to listen sounds from heart, lungs with stethoscope, as a part of medical diagnosis.
■ Analysis of respiratory sound using stethoscope depends on physician experience, learning and ability to recognize and differentiate pattern.
■ One of the recent study in the Journal of American Medical Association, examined the stethoscope skill of various kind among 453 practising physician and 88 medical students. Whatever their age or experience, doctor correctly recognized 20 % of disease.
■ All this possess a need of design of computerised expert systems to assist physician, or sometimes even before the arrival of the physician, as a first aid tool for disease detection, with maximum accuracy and minimum cost.

Description:
The system is divided into two main phases namely:
1. Training Phase
2. Testing Phase Training Phase:
Since it is an intelligent system it is important to give training to the system. Here in the training phase the following phases are implemented.
a. Sound Input: Here the input of the sound is given to the system (speech
recognition).
b. Preprocessing: Here with the help of LMS the system will find the sound
properties.
The LMS algorithm:
Input:
Tap -weight vector Input vector Desired output: d
Output: Filter output: y Tap -weight vector update:
1. Filtering: y
2. Error estimation.
3. Tap-weight vector adaption.
c. MFCC: IN this phase the Mel Frequency components are extracted from the
sound.
d. Disease diagnosis: the features of the sound for the normal and infected are
stored in the database of ANFIS.
Testing Phase:
In this phase the system will be tested by the experts to test the output of the system.
The experts can be doctors from the same field. Here in the testing phase the
following phases are implemented. __ _ _.
a. Sound Input: Here the input of the sound is given to the system (speech
recognition).
b. Pre-processing: Here with the help of LMS the system will find the sound
properties.
c. MFCC: In this phase the Mel Frequency components are extracted from the sound.
d. Feature mapping: The extracted features are mapped with database coefficient
values by ANFIS.
e. Output: Output is generated as asthmatic or non-asthmatic following the adaptive
rules of the system.

Claims:
1. An Asthma Detector Using Fuzzy Neural Approach having following essential
requirements:
A physician or patient first aid assist should have a Lap top /computer with voice recording facility and Matlab software installed in it. The detector module will be installed and Trained and tested. For this Create database (DB) saves all the entries as given by the user at input. Once DB is created, the ANF1S is trained. After training is completed, Evaluate DB is applied to ANFIS classifier to get the required class. The graphical User Interface is developed and the corresponding input audio wave file and the MFCC plot, for asthmatic and non-asthmatic samples appears. The database of around 1000 samples is used. It is procured from the labs of renowned Pulmonologist. This is accessed randomly for training and testing of the ANFIS system. Real time training and testing is carried out using recorder on Laptop.
2. The detector apparatus of claim 1 wherein said computer includes one display,
MATLAB software, detector program, a recorder and a speaker.
Thus we claim an indigenous Asthma Detector Using Fuzzy Neural Approach with following advantages
3. Real time training and testing using recorder on Laptop.
4. The application developed is user friendly and can be embedded as a mobile App.
5. The System consistency with ANFIS system using MFCC as feature extraction.
6. A fast, non- invasive, economic, easy to use first aid tool for asthma detection, even
before the arrival of physician.
7. The system is accurate with average accuracy mounting to 95.77%. It can be majorly
used by the doctors for primary detection of the disease. It will save a lot of hard work and time of the experts.

References:
1. Vimala, C.Radha, V.: 'A review on speech recognition challenges and approaches', World Comput. Sci. Tnf. Technol., 2012, 2, (1), pp. 1-7
2. Anusuya, M., Katti, S.; 'Front end analysis of speech recognition: a review', Int. J.
Speech Technol., 2011, 14, (2), pp. 99-145 environments using wavelet transform'. 2002. Available from: ttp://www.wseas.us/elibrary/conferences/skiathos2002/papers/447-, 231.pdf
3. Morgan, N.: 'Deep and wide: multiple layers in automatic speech recognition', IEEE Trans Audio Speech Lang. Process., 2012, 20, (1), pp. 7-13.
4. O'Shaugnessy, D.: 'Interacting with computers by voice: automatic speech recognition and synthesis', Proc. IEEE, 2003, 91, (9), pp. 1272-1305.

5. Ranjit Sawant, A, Ghatol, "Neuro-Fuzzy Approaches for Analysis and Diagnosis of Respiratory Infection Diseases : A Survey" published in Ciit International Journal of Biometrics and Bioinformatics, Vol 8, No 01, January 2016.
6. Ranjit Sawant , A. Ghatol, " Fuzzy Model for Diagnosis of Respiratory Infections using Respiratory Inductance Plethysmography Data" published in 2016 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS).