Panasonic AI | Panasonic Research and Development on Artificial Intelligence (AI)

High Precision Real-Time Pedestrian Detection Technology

Although AI is capable of performing highly precise recognition and evaluation, it requires an enormous volume of calculations to perform these functions. Furthermore, to be able to commercialize pedestrian detection technology for application in automated driving, AI developers need to address the key issues of reliably detecting pedestrians in various postures, while also reducing processing time and conserving energy to make AI technology mountable in devices and systems.

Existing Algorithm

Conventional method for High Precision Real-Time Pedestrian Detection Technology. Two people crossing at a crosswalk to a parking appear on the monitor (photo). First, the candidate of pedestrian detection are is extracted (existing image recognition technology). Second, processing is performed to detect pedestrian for extracted area (recognition processing by repeated deep learning). Finally, only one on the left of the two people  on the crosswalk is captured.

Developed Algorithm

Newly developed method for High Precision Real-Time Pedestrian Detection Technology.  Two people crossing at a crosswalk to a parking appear on the monitor (photo).  Next, the deep learning model is used to map the likelihood of pedestrians from the entire image (probability distribution map). Finally, both pedestrians are captured on the crosswalk.
Supplementary photo for the newly developed method for High Precision Real-Time Pedestrian Detection Technology. A pedestrian has been captured between cars in a parking.

Panasonic is researching and developing AI technology for realizing real-time pedestrian detection, while maintaining a highly accurate detection rate. To achieve this, we are leveraging our expertise in mountable systems and applying our in-house-developed pedestrian detection algorithm based on Deep Learning.

Coordinated control technology for people and robots

Technology that can safely coordinate and manage people and robots is essential for us to be able to use robots in our daily life.

Self-Reliance Support Robot

Panasonic is developing a Self-Reliance Support Robot to ease the mental and physical burden on the people who give and receive care in our aging society, and to assist the elderly in their daily life, such as getting out of bed and moving to the toilet, as well as sitting on the toilet seat.
Using Panasonic's original coordinated control technology for humans and robots, the Self-Reliance Support Robot senses the intentional movements of an elderly person when he/she wants to stand up, and provides just the right level of assistance, allowing them to use their remaining muscle strength and preventing unnecessary muscle atrophy or strain.

Photo of a woman tryint to stand up using the Self-Reliance Support Robot(exhibited at the International Home Care & Rehabilitation Exhibition 2014 and 2015)

Guided Walking Support Robot

Panasonic is striving to help elderly people live a longer and healthier life by providing assistance in carrying out daily activities and helping them to live a full and happy life.Our research and development focuses on enabling elderly people to live an active daily life physically, intellectually and energetically, using robots that carefully sense the physical condition of each person and interact with them at the optimal level.
The Guided Walking Support Robot predicts the intended walking direction of the person from their grip on the handle and includes this information in their direction and strength of guidance, allowing the person to move safely and freely around the house.

Photo of the Guided Walking Support Robot and another of a woman using the robot (exhibited at Japan Science and Technology Agency Fair 2015 in collaboration with Nagoya University COI)

Interactive and Machine Translation Technology

The growth and spread of the Internet is connecting people around the world, as well as increasing their opportunities for communication in the real world.
To achieve deeper and closer communication with people overseas, we need to eliminate the language barrier that still exists in Japan.
To realize this, Panasonic is developing machine translation technology primarily for work involving human interactions, so that Japan can provide even better services to overseas visitors.
Our developed machine translation technology automatically generates various types of expressions, such as responses to a range of usage scenarios. The machine is also self-teaching, which progressively reduces the frequency of errors and improves translation performance.

Enhancing the Translation Engine using AI

Panasonic is applying AI to generate the corpus that forms the basis of its translation technology. AI technology is used to analyse and identify existing translation examples: it then automatically generates various patterns of expression. This will enable a more natural translation of dialects and everyday conversation as well, while automation will help to accelerate the advancement of machine translation.

Illustration of an Asian person and a Westerner conversing in Japanese and English, respectively

Prototype of a multilingual voice translation machine for a range of usage scenarios

Photo of the multilingual speech translation device prototype