An Embedded Speech Recognition System for Children

Kaveh Azartash, Ph.D., CEO and Dhonam Pemba, Chief Scientific Officer, KidSense.aiKaveh Azartash, Ph.D., CEO and Dhonam Pemba, Chief Scientific Officer
The onset of computer-assisted learning has led to the rise of speech recognition systems, allowing people to enjoy learning through voice-enabled and adaptive technology. The global voice recognition market is expected to reach $127.58 billion by 2024, according to a new study conducted by Grand View Research, Inc. The increasing number of benefits offered by the voice recognition technology, including easy accessibility, better productivity, and flexibility are anticipated to drive the demand for speech recognition solutions over the next eight years. The increased demand for high comfort and convenience is also expected to propel the demand. However, a child communicating with technology (software or hardware) is a different ball game, for which, the industry is ill-prepared for. The inability to accurately recognize kids’ speeches in the regional accents is a key challenge to the market. Additionally, with current technology platforms driven by cloud computing and cloud storage, the sensitive information about the kids is in jeopardy. With very limited compliance with the necessary regulatory requirements, the online speech recognition systems pave the way for bigger corporations to use kids’ data for other commercial purposes, thereby making kids data vulnerable online. In these circumstances where there is a threat to the integrity of kids’ data, Irvine-based Kadho has developed a highly COPPA compliant and efficient embedded speech recognition system,, to help technology understand kids. With an inbuilt artificial intelligence (AI) engine combined with neuroscience of early language acquisition, Kadho’s offline speech recognition system works at the device level which means data does not get transmitted out to the cloud-based servers. Compatible with a variety of devices including smart speakers (such as Amazon echo and Google home), wearable, robotics, and toys, strictly abides by COPPA and offers the same services which cloud-based speech to text platforms do, but with improved latency and security.
Applications of speech recognition for children go far beyond entertainment. Children today are utilizing technology for educational purposes from language learning to other subjects such as science and math. Speech recognition has vast potential for these applications for children. It is very essential to note that speech recognition software developed for adults will not be useful for young children. Having a database that is built on speech data from over 150,000 native speakers from different countries, allows children to communicate with technology in a highly secure manner. Currently, KidSense. ai platform supports eight languages, including English, Mandarin, Japanese, Korean, Spanish, French, Italian, and German. “Despite having a substantially large language acquisition database, Kadho regularly adds information to the database to refine the system,” says Dr. Kaveh Azartash, CEO of Kadho. The platform has a latency rate of 100 milliseconds, making it fast, accurate, reliable, and convenient for users. “Our speech recognition system can be customized according to the device on which it is to be installed.”

Our goal is to provide a secure and COPPA compliant platform for children to enjoy talking to technology around them

With a technology that understands kids, has been successful in establishing another educational-focused technology to quantitatively assess how kids speak in real time. leverages its unique artificial intelligence based speech assessment algorithms to produce data around pronunciation, pace and intonation of kids’ speech and provides real-time feedback for children, educators and parents. To ensure that its solution is validated and compatible on various web and mobile applications irrespective of device hardware configurations, Kadho has partnered with six Asian robotics companies and several third-party educational companies.

Having revolutionized the paradigm of kids’ speech recognition, is now looking to go beyond the consumer-facing products. “In the months ahead, we want to use our offline speech recognition system to partner with chip manufacturers to develop a unique and secure solution to handle kids speech for a variety of industries,” concludes Dr. Azartash.