Projects
ACTS
ACTS is a python based framework that specializes in multi-device testing. It is mainly used to test Android Connectivity and performance.
ACTS Telephony - click here
Code Repository - click here
Tools: Python, Java, Google Cloud Platform, GNU Common Lisp data, Shell, JSON
IMPlantable medical device
In today’s world, Gastrointestinal (GI) disease is one of the common problems in human body. Gastrointestinal disease mainly refers to gastrointestinal tract namely the oesophagus, stomach, small intestines, large intestine and rectum. According to National Institute of Health, most people have experienced gastrointestinal disease in stomach. Gastric peristaltic contractions are responsible for mixing and emptying food contents that are indigested throughout the day. Various gastrointestinal chronic disorders such as gastroparesis and functional dyspepsia, affect the gastrointestinal motility. It has been hypothesised by many researchers that the spike activity signals of the stomach are responsible for triggering peristaltic contractions. The slow wave of the gastrointestinal tract mainly functions to trigger the contraction of smooth muscles. These slow waves are generated by Interstitial Cells of Cajal (ICC). Interstitial cells of cajal are the thin layer lying between circular and longitudinal smooth muscle cells layers. Oscillations in membrane potentials generates slow waves. Both circular and longitudinal muscle cells are electrically coupled to the network of ICC and thus slow wave propagates with small change in waveform over gastric tissue. These electrical waves originate from the upper corpus of the stomach and travelling around the circumference of the stomach towards antrum. With the aim of detecting and analysing slow wave, a person can easily determine the behaviour of the stomach.
A multi-channel recording, and stimulation system was designed to study human stomach motility in a closed loop structure. The system consists of three subsystems namely implantable unit, wearable unit and stationary unit. The implantable device has a recording unit and a stimulation unit. Recording unit is used to record gastrointestinal electrical activities from multiple electrodes in the stomach. Stimulation unit is used to apply stimulation pulses to the electrodes placed in stomach. Stimulation pulses are applied for smoothing the motility of the system. The wearable system was tailored for maintaining the communication with the implantable system and stationary system. The modules are communicating within the ISM band set for wireless communication. Stationary system give access to the user to monitor real-time gastrointestinal electrical activities and act according to the variation of motility in the stomach. A graphical user interface (GUI) was designed in LabVIEW. The GUI system will acquire gastric electric activities wirelessly from the implantable device in stomach. The main function of the GUI application is to perform real-time and off-line analysis. GUI designed ICC network based on finite state machine principles. The model compromises three super states namely pacemaker, corpus and antrum. A network of ICC was proposed that consists of eight super state representing 8 rings on the stomach. Each ring consists of four sub states representing channels. The rings are located around the circumference of the stomach. Two rings are situated in the pacemaker area i.e. fundus of the stomach. Similarly, four rings in corpus and two ring in antrum. When a slow wave is generated in each channel, it activates each super-state subsequently whole ring. Each rings are placed 4mm apart. GUI application should verify the data integrity of the command packet and then extracts the information. GUI application performs digital signal processing for improving the performance of the system by calculating the time instances when the slow wave is generated in each channel. The multi-channel system was successfully validated with laboratory experiments. The user has the ability to set a dynamic range of frequency for the treatment of various gastrointestinal diseases. The system has the ability to apply stimulation pulses to the corresponding electrodes when the frequency does not correspond to the dynamic range assigned by the user. The model simulated identifies three patterns namely normal, bradygastria and tachygastria. An input applied to a single ICC activation, the model responds with normal antegrade activation of 3 cycles per minute. The propagation speed varies in pacemaker, corpus and antrum depending on the condition the model is stimulated. For tachygastria the model stimulates 5 cycles per minute and for bradygastria the model stimulates 2 cycles per minute. The system is capable of delivering electrical pulses at amplitudes in the range of +12mA to -12mA to the load of 1k ohm. The chain of pulses applied to a particular ring demonstrated effective restoration condition from bradygastria to normal. Hence, the presented model shows the effectively stimulates normal rhythmic and dysrhythmic of gastric activities in real-time. The model is also computationally efficient to analyse the data in real-time application.
Key Tools: Altium Designer, LabVIEW, PyCharm, MATLAB, Code Composer Studio
Programming Languages: C, C++, Python
Group No.7B Android Based App For Emotion Recognition From Speech.pdfAndroid based application for emotion recognition from speech
With the aim of recognizing not just what a person says, but also with what intention the words are being said, we have designed an Android Application for emotion detection from speech. This is due to the fact that even non-verbal communication says a lot about the intention of the speaker. In addition to spoken words, the manner in which the words are spoken tells us a lot more about the speaker's intention. The same message or words can be conveyed with different semantics (meaning) by incorporating appropriate emotions. For example: the word “OKAY” in English is used to express admiration, disbelief, consent, disinterest or assertion. We have used Android powered smart phones in our project due to their obvious proximity to the users and hence, subsequently to their emotions. The emotions to be recognized from the user's speech are to be broadly classified into the following 7 types: Anger, Boredom, Disgust, Fear, Happiness, Sadness and Neutral; the voice samples for each emotion are available in the Berlin database called Emo-DB and these voice samples are used in our project. Our project basically involves: developing an Android App, extraction of features from the acquired speech and then the classification of these features based on certain parameters. We made use of the Android Studio IDE which is official integrated developing environment developed by "Google‟ for App developers for developing on Android platforms and also provides some predefined classes. We used Audio Record and Audio Track classes and their subsequent methods for recording and playing speech. Then speech features such as Pitch, Energy, Formants, MFCC (Mel Frequency Cepstral Coefficients), Pitch difference and Energy difference are extracted from voice samples of the Emo-DB database as well as the input speech. Then using a classifier, these features are classified into different emotions. We also take into consideration the adverse effects that noise could have on emotion detection in case the input voice is affected by noise and hence, to lower its effect we make use of the Noise Suppressor class in Android to avoid interference from stationary as well as non-stationary background noise to improve the accuracy of our application.