Pune: Have you ever wondered why investigation agencies take time to solve complex cases? One of the reasons is the lack of tools and technology to identify and validate any suspect’s speech, text or video sample from the recorded sample and circumstantial evidence received. For example, a central investigation agency asks a suspect to make 300 signatures just to ensure that it can identify the “expert forgery.” After getting those signatures, they manually verify all the samples to reach a conclusion. The same goes for any sample of recorded audio evidence from a suspect, where officers “listen” to the audio multiple times to conclude whether the evidence and the recorded samples are from the same suspect individual.
Retired teacher-turned-entrepreneur and founder of Anubhooti Solutions, 69-year-old Shaila Dinkar Apte has designed and developed tools that help in speaker verification (suspect’s voice sample) through word comparison, audio segregation of different speakers, audio authentication (in case of manipulated audio), emotion detection, noise cancellation, signature verification and document analysis and image authentication (in case of deeply fake images).
Shaila, a passionate researcher hailing from Thane, who was denied admission into a research institute for not meeting a certain criteria, pursued her dreams despite several domestic challenges. She began “coding” in 1980, when her visionary husband, a former IIT Bombay employee, brought a “personal computer” to her home. Shaila completed her Mechanical Engineering (ME) and her PhD (Electronic Engineering) from Walchand College of Engineering, Sangli.
Shaila said, “As a professor of electronics and digital signal processing, I worked at Walchand College in Sangli for 28 years and at Rajarshi Shahu College of Engineering in Pune for 12 years from 2007. I was very interested in student projects and used to enjoy working hard for any innovative problem. Before I retired at the age of 62, I had registered my own company called ‘Anubhooti Solutions’. I was guiding PhD candidates in my field on topics like speech improvement, writer verification, handwritten character recognition, classification of speech disorders, etc. and even in other fields that I was passionate about. I have published five books with international publishers in the area of signal processing. This enriched my research experience with a team of 11 successful doctoral candidates. I guided more than 60 ME candidates working primarily in the speech processing domain with challenging research problems such as emotional speech, audio signature, audio watermarking, verification and speaker recognition using many different innovative ideas.”
“While helping students, I wrote several books widely accepted by the industry. The founders of a Delhi-based startup approached me to help them develop their voice-based product to generate stories with the voice of parents. I managed to make a prototype in six months, but since the founders did not have enough lead, product development was stopped. But this short period gave me confidence about the industry’s prospects. I had extensive research experience, but I didn’t have the experience running a company and developing commercial products,” Shaila said.
Shaila received another project from a Canada-based company to work on “emotion detection” and “speaker diarization” from voice samples. However, she was deceived by the company that took away the software code in exchange for only $2 lakh.
“I had never looked at the product from a money-making perspective and had therefore forgiven the Canadian company. Meanwhile, another entrepreneur, Suhas Kulkarni, director of Lab Systems and Biotech (India) Private Limited, who worked for forensic laboratories and had seen the problems encountered by the agencies, approached me to develop some tools for audio analysis.
“Suhas and his friends had also worked in the speech domain earlier. As part of their research, they found my book and thought I would do the job better. They asked me to generate a tool for audio analysis. I decided to do it because I had almost 40 years of experience in the field of speech. Relying on my practical knowledge, I decided to do the best I could.
“Meanwhile, the Covid-19 pandemic hit us and the lockdown was announced. We used this period of approximately three years to develop the entire range of speaker diary, speech laboratory and speaker verification products ready for the market. We met only once before the lockdown, but afterwards we had continuous interaction to generate finished products. We started the work with a small team. Our only lack of experience was the operation of the company and its related jobs,” Shaila said.
“There are around 200 forensic science laboratories (FSL) in India. These institutes train and educate in forensics, analyze audio clips captured by investigative agencies and need tools for analysis. Currently, there are some tools available for such analysis in forensic applications, but they are not user-friendly and difficult to operate. They have also not been found to obtain accurate results. Therefore, good, accurate and easy-to-use tools are needed. We took it as a challenge and developed six products for the use of audio forensics. We believe that these made in India products will be useful for the society,” he stated.
Go to the market
Anubhooti Solutions has collaborated with Lab Systems and Biotech (India) Private Limited and M/S Lab Systems (I) to market their products in India and across the world. “Until now we have focused solely on the government sector and our clients are laboratories in the forensic field,” Shaila said.
Revealing their future development plans, Shaila said: “We are receiving requests from FSL to generate a special system dedicated for the detection of AI generated audios, images and videos. Additionally, agencies also need a stress detection feature along with emotion detection. We are working in this area and hope to launch a solution by the end of this financial year. We have Anubhooti signature verification, document analysis and image authentication products in the image analysis domain and video authentication will be launched soon.
“We have a 2 to 3 percent margin of error allowed. Although we have automated all processes, the last call is always taken by an experienced analyst. Our tools will never replace them, but they will help streamline processes. Conclusions drawn from reports or analyzes of our tools are presented and admissible in court.”
No language barrier
Shaila said, “All the tools developed by Anubhooti Solutions are language independent and can work in any Indian and foreign languages. Therefore, it is useful all over the world.”