mansoor@ccs.neu.edu
pervaiz.m@husky.neu.edu

Download PDF

Download Jottings

Hear Me (Java on Microsoft Surface)

People with speech impairments have a hard time communicating with the general public. For Dysarthria (a class of neurological speech motor disorder), in the beginning patients are able to communicate with people because they can understand them (though with difficulty). However, as the impairment progresses the communication becomes labored and eventually the patients reduce the communication attempts. As a result, the network of the patients shrinks considerably. So much so, that only the caregivers of the patients are able to understand them for a very limited set of utterances. This reduces the self-reliance and autonomy of these patients.

There are no demographics available for patients suffering from various kinds of dysarthria, but Parkinson's disease, ALS and MS are just some of the conditions where dysarthria is a symptom. This means that there are more than 5 million patients in the US alone.

This application will be a simple recognition system, which will take the patients' utterances and generate comprehensible speech for it. However, this communication will be similar to the communication between the patient and the caregiver, where the caregiver can only understand a very limited set of utterances of the patient. The caregiver will train the system by recording the patient's utterances ten times for each output sentence. The system will then be able to listen to the impaired speech utterance from the patient, identify the matching sentence and then synthesize the speech for it. There might be a limitation in the number of utterances the system can recognize so as a further addition, the sentences will be grouped into categories according to situations (for example a doctor's clinic, or a restaurant). The patient will choose the category before speaking. This will allow the system to recognize larger number of utterances.

There have been efforts to build a free-speech recognition system for Dysarthric speakers but with little success. However, this is not a speech recognition system per se. This is a much simpler application, which will only endeavor to recognize a limited number of sentences.

Technologies: Java

Assist Your Friends (Speech Based Application- IVR)

When a patient is diagnosed with diabetes, the medical professionals pile up a quite a lot of information on the new patient. This knowledge transfer comprises information about the disease, medical regime, dietary restrictions, exercise requirements and so on. This overload of information over-whelms the patient who is already stressed with the breaking of the news. These patients always have follow up questions. Someone who is already suffering from the illness can answer most of the questions based on experience. This may result in taking load off the clinicians. One other potential benefit is that the patient may ask more questions from a peer rather than an authority figure (doctor).

The solution is Assist your friends which will be a phone application to help diabetics with better control of blood glucose. It will be a peer support system (like patientslikeme.com but over a phone) where diabetic patients can call in and ask questions, while other people who are experienced diabetic patients can listen to those questions and record answers to assist in glucose control. Other experienced patients can further improve these answers.

It is different from patientslike.com (which already serves patients with diabetes) as this is not an online system. This is a phone service where the users will call a phone number and record and listen to queries and answers. It is important to diverge from a web based online system because 47% of adults above 65 do not use Internet and 58% of the people earn under $50,000 don't have smart phones. Therefore, a web based system or smart phone app is not useful for supporting the underprivileged population or the population that just doesn't use Internet.

Technologies: Tropo (with PHP or Python), Java

Better Living (Facebook App)

A large population in the US suffers from diabetes, hypertension and heart problems. Despite the fact that many people understand that improper food intake, not exercising regularly and not taking proper sleep goes against their daily blood glucose and blood pressure values and puts them at further risk for organ failure, heart attack, stroke and brain hemorrhage. Still they are not careful about these other factors besides following a medical regime, which is very crucial. People who maintain a healthy life style after being diagnosed with such medical issues are able to live a normal life with lower doses of medicine and fewer complications.

Augmenting to the lack of health lifestyle, at times people live alone, away from their close family and friends, so these possible influences (family and friend) cannot impact the person in following the proper diet, exercise and sleep regime. There is a strong need for a system, which informs the close influences of the user in cases where he/she would need support and encouragement and motivation to take (or continue taking) steps for a better living.

Better living is a facebook application where the users will get support from their close family and friends in times of need. The application will allow the user to input information about daily:

Based on the values of input variables over a certain time period, the application will post notifications to the user's friends and indicate a possible concern. For instance, if a user has not exercised in the last three days, and the sleep pattern of the user is erratic, the system will raise an alert to the friends and family about the degrading trend that could possibly affect user's health. The family member or friend will view the variables, which cause the alert and will be able to communicate with the user. Since, they will know the user in question, they will be able to provide better motivation and support.

Technologies: PHP or Java