EZSpeech

Problems: 

1) Shortage of SLPs – currently there are 400-800 kids per one speech pathologist (SLP) 

2) High fees are not affordable by many families – Fees per session is ~$60 for on-line, ~$120 for off-line 

3)  Low efficiency of the treatment - Normally SLP spends with a patient~30 minutes twice a week , while effective treatment requires daily practice. Parents rarely make sure that children practice daily

Our solution

We are developing an APP which will make speech therapy available to underprivilaged families. Our fees are just $3-5 per session. 

In our App, AI as a virtual SLP will guide a child throughout the exercises and assess the accuracy of performance using computer vision. App covers mimics, voice and gestures. It'll generate stats what and how the exercises were performed as well as progress reports.  

We will gamify the APP to make the process interesting for children.

Currently in our Speech Center we serve around 200 patients a year, including 180 from low-income families. With this application we will be able to increase number of patients 10 times. Furthermore, we will be able to serve kids from other states in USA, especially those who couldn't afford our standard fees for off-line and on-line.

There are over 40 million children worldwide suffering from speech disorders. We can help not only kids from US, but also children from other countries in Asia, Latin America and even Africa. Because the App visually demonstrates an exercise and provide native guidance so a non-English speaking children can use it

The App processes and interprets combinations of contractions of muscles of face, voice, tongue and  lips simultaneously. 

Transfer learning from SOTA + production approved

Technological approach is based on:

a) state-of-the-art artificial neural networks

b) production of approved neural networks and ensemble architectures

c) transfer learning to the subject area

Specific datasets + scalability by architecture

The architecture is made to scale. i.e. covers new cases within the subject area so that their development is cheaper and faster than from scratch

A dataset is prepared that is specific to the subject area. It has value in itself, because it is not just a video, but marked-up frames with measured reactions of the neural network: expected vs fact

We follow a UN principle "Leave no one behind"

Correct speech is a crucial element of any educational process ensuring ultimately both social and economic integration essential for the well-being of both individual and of the entire society.

Due to high cost of the treatment and the fact that speech therapy is not always covered by insurance, a lot of families simply do not have access to speech therapy.

Our target audience includes:

1) children and adults from low-income families, who can not afford SLP. After piloting the App in our Center we plan to expand in US and worldwide, primarily in Latin America, Asia and Africa

2) children with cerebral palsy and autism - We hope gamification of the process and innovative approach will drastically increase their engagement and make a boring and difficult process a little less challenging for them. We have already discussed and agreed a pilot with a boarding school for children with cerebral palsy

3) non-English speaking children and adults (in U.S. and worldwide) - in our Center we serve immigrants. Many of them don't speak English and don't always understand explanation of SLP.  Visual demonstration of an exercises as well as step-by-step guidance does not require knowledge of English. So non-English speaking patients can be treated worldwide. 

1) Support SLP - our APP is a simulator for exercises prescribed by SLP to increase the effectiveness of the treatment (within the Speech Center or at home)

2) Personalized instruction - AI tailors exercises, instructions, guidance and assessment to specific needs of a specific kid. The algorithms are tailored to children with different gender and ethnicity to ensure equal quality of mimic recognition and interpretation

3) Communications - progress reports and stats provide objective views on performance of a particular exercise or set of exercises. Computer vision can unambiguously detect misperformance. Progress reports are visualized for parents convenience. 

We set up and trained the first tier of the neural network to recognize 30 key-points of face of any age of ethnicity. The neural network detects detailed muscle activities The model is interpretable, but not in automated mode yet. Next stage is to develop a second tier of the network to automate interpretation + automate markup. We plan to run pilots in our Speech Center starting from August 2021.

According to our plan the pilot for face and cheeks is scheduled for end July-August, tong, teeth and lips - end September-October, voice-November and fingers - December, other gestures - January.

After internal pilots within our Speech center we have agreed to run pilots in a boarding school for kids with cerebral palsy in Novosibirsk (Russia) (letter of intent has been signed)

Presently Speech therapy services are represented by off-line and on-line sessions (by a human SLP ).  Our solution offers a virtual SLP (on-line simulator), which enables kids to practice independently without supervision of a SLP or a parent. 

Similar technology is used in apps for yoga, fitness or back exercises (for example https://www.kaiahealth.com/  ). But no one has developed a software for speech therapy, especially multimodal.

In terms of innovations, we have two:

1) tracking and interpretation technology exists, but for speech therapy has not been yet developed. We created a team of IT developers and experienced SLPs to develop algorithms specifically for the purpose of face muscles, gestures and voice

2) multimodality - to the best of our knowledge the existent software covers maximum two modules, where is our will combine three: face, gestures and voice

New approach to the problem:

1) speech therapy will be available for kids from underprivileged families

2) effectiveness of treatment will increase as kids will combine their practice with human SLP with daily exercises through our App

3) hybrid-therapy and Web-simulators - can be used to optimize the work-load on SLPs and operating expenses of Speech Centers and Hospitals

4) objective assessment of progress generated by our App can be submitted directly to SLP or medical records to analyze the effectiveness of treatment   

Currently our solutions is not available to public. 

The solution will be launched in September in two countries - U.S. and Russia in Speech Center in New York and Boarding School for kids with Cerebral palsy in Novosibirsk (Russia) - in total around 400 kids.

Within 1 year we plan to reach 4-5 thousand users

Within 5 years we plan to reach at least 5 million users worldwide, primarily in USA, Latin America. Africa and EMEA region 

1) we developed a prototype and found an investor to develop MVP and final product

2) we developed a 12-month schedule and so far we are on it

3) we found three entities where we pilot our MVP 

4) throughout the entire process we discuss the App functions with potential users to receive their feedback (specifically the order of our exercise plan is based on preferences indicated by school for kids with cerebral palsy)

5) we reached preliminary agreements with two foundations to run pilots in schools and speech centers for underprivileged kids  

6) we have a short-list of companies, schools and organizations we contact as soon as we have piloted MVP (in US, Spain and Russia)

3 people are working part-time. This is until we start piloting. Since the pilot and after launch they will work full-time

3 people are working full-time (IT developers)

Adelina Samuels is a Qualified Speech Language Pathologist with over 20 years of experience. She is an Active member of ASHA (American Speech and Hearing Association). She is a Clinical Director and Founder of Medical Group Little Steps. All her professional career Adelina have been providing serviced to low income patients with speech and language disorders and disabilities, as well as army personnel and their children that had suffered from neurological and cognitive disorders and disabilities. Those services are provided through state organizations and medical insurance companies.

Denis Bondarenko has nearly 20 years of experience in IT industry. Since 2001 Denis has been engaged in computer vision, 3D technologies, and video data enrichment.  Denis's son has speech disorder so Denis is keen to develop an App to help his son and other children
 

Firouza Eshonova has over 20 years of investments, assets management, IPO and M&A experience (including Big 4 in Moscow and London). She is ex-CFO of one of the largest industrial groups in Russia and independent consultant to international commercial companies. Firouza's son has suffered from severe speech problems and received a long-lasting treatment in Speech Center. That's why she come up with the idea for this project and created this team.

Our goal was to create a team of highly qualified professionals who:

a) understand the problem and the target audience 

b) can execute technological solution and 

c) can execute the project based on principle "impact for profit", i.e. combination of business-for-profit with the main goal to help those who is in need and to achieve the greatest impact 

We didn't specifically search people based on their gender or color. Nevertheless, 2 of 3 our leaders of women and both are minorities. Adelina is a refugee and Firouza is Asian. Both Adelina and Firouza are self-made women with outstanding professional career.

 Our main and only criteria when we searched for investors was to have an "Impact investor", who's goal is not limited to profit generation but who places a high importance to SDG. Our current investor is a philanthropist who introduced us to various organizations who help children with disabilities.

Our goal for the next 12-24 month to develop partnership and alliance with Charities and Foundation, as well as impact investors (we are discussing a membership in our Board with a couple of representatives).

Ideally, we would like to have representatives for various groups (black, latino, kids with disabilities etc)    

Funding is not the main purpose of our application. We will be happy to join the program even if grant money isn't available. 

The main two purposes we are applying to Solve are:

1) Access mentorship, coaching and strategic advice from experts, as well as the Solve and MIT network

2) Join a powerful network of impact-minded leaders across industries and sectors

SCALE IMPACT in priority - Reach out those who is in need 

We have a network to access low-income communities in NYC, NJ and several cities in Russia and Uzbekistan. But we would like to reach out low-income communities in Latin America, Asia, Africa etc.

So we need assistance to get access to specific organizations (schools, speech centers, rehabs etc) in those countries, especially those from low-income communities. We are confident that through SOLVE network we can find people who can introduce us to those communities.

SCALE IMPACT (diversification) - Adaption of our technology to other segments 

Based on our interviews with professionals we prepared a further plan of development. As such: 1) release of similar App for adults (after CVA or for alzheimer’s disease) 2) development of additional services

relevant to speech therapy such as neuropsychological exercises and
set of exercises for kids with autism  3) connection of the software to VR glasses and exoskeleton etc

So we will be happy to learn more how we should diversify our technology to achieve a larger impact 

Human Capital

We will be honored to have someone from SOLVE in our Board as independent Director(s). We also are looking for Business Developers in Africa, Latin America and Asia. 

Business Development (PMF and strategy)

As mentioned in the previous section we plan to diversify our technology to other segments (adults, kids with autism etc). We are constantly talking to various groups of potential users to further develop our software and increase customer satisfaction

Product/Service Distribution

There are over 100 million kids with speech disorders. Many of them can not afford speech therapist. We can't help everyone who is in need at the moment (at price below our standard fees). So we need to prioritize distribution of subsidized subscriptions and licenses. We have a list of potential recipients of subsidized services and will be happy for any advice and assistance to expand this list further. 

Public Relations

To raise awareness that very cheep speech therapy service is available (whether subsidized on standard rate which is still 10 times lower than market)

Legal and Regulatory matters

1) We pay high attention of compliance with local rules and regulations (including but not limited to personal data protection). Therefore, from time to time we get legal advice.

2) Our medium-term plan is to register our software for medical purposes (support for medical-decision support). It is crucial to ensure we comply with all applicable rules and regulations in specific country  

1) Schools, Speech and Rehab Centers, Hospitals 

2) Charities and Foundations

3) Individual professionals who can help us with the areas described above

1) Our solution is digital and innovative (computer vision, AI and machine learning)

2) The most important is that our solution gives kids with speech disorders confidence that they can do it - they can improve their speech and get rid off disorders they have. Self-confidence is critical for further professional and personal development. And we give it in an interesting gamified manner involving kids in a process. 

3) Our solution addresses current problems of inequality, because we give colored kids a chance to access speech therapy at affordable rates.  

4) It is tailored to US-based primary classrooms that provide experimentation (kids are introduced to a virtual SLP, we are sure that they not only improve their speech but also get interested in AI technology). 

Our solution qualifies for the AI humanity Prize for the following reasons:

1) Our software will improve the lives of individuals and our global community through neuroscience research and information technology

2) we use computer vision, AI and machine learning, as well as engage kids through gamification

3) We disrupt the market by making speech therapy available at price of $3 vs $60

4) Our solution will provide objective stats and progress reports on treatment (unlike subjective and at times vague assessments from a human SLP) 

We are proud to give kids around a world a chance to improve their well-being.

See section above for qualification

We can use the Prize for several purposes (depends of size of the prize):

1) expanding our technology to other segments. As mentioned above we are planning to develop a version for adults (after CVA or alzheimer’s disease) and kids with autism

2) connect our software with VR glasses and/or exo-skeleton to increase the outcome and kids involvement 

3) to issue subsidized licenses to speech-centers in low-income communities