论文部分内容阅读
感知是智能系统与现实世界的交互界面。如果没有复杂而灵活的感知能力,就不可能创造出高级的人工智能(Artificial intelligence,AI)系统。最近,潘云鹤院士提出了AI 2.0的概念,其最重要的特征就是未来的AI系统应拥有类人甚至超人的智能感知能力。本文简要回顾了不同智能感知领域的研究现状,包括视觉感知、听觉感知、言语感知、感知信息处理与学习引擎等方面。在此基础上,论文对即将到来的AI 2.0时代智能感知领域需要大力研究发展的重点方向进行了展望,包括:(1)类人和超人的主动视觉;(2)自然声学场景的听知觉感知;(3)自然交互环境的言语感知及计算;(4)面向媒体感知的自主学习;(5)大规模感知信息处理与学习引擎;(6)城市全维度智能感知推理引擎。这些研究方向应在未来AI 2.0的研究规划中进行重点布局。
Perception is the interface between the intelligent system and the real world. Without sophisticated and flexible perceptions, it is impossible to create advanced Artificial Intelligence (AI) systems. Recently, Prof. Yunyun Pan proposed the concept of AI 2.0. The most important feature of AI 2.0 is that future AI systems should possess intelligent perception capabilities of humans and even superhuman humans. This article briefly reviews the research status of different areas of intelligent perception, including visual perception, auditory perception, speech perception, perception information processing and learning engine. On this basis, the dissertation looks forward to the key directions that need to be vigorously studied and developed in the upcoming AI 2.0 era, including: (1) active vision of humans and superhumans; (2) perceptual perception of natural acoustic scenes ; (3) speech perception and calculation in natural interaction environment; (4) autonomous learning for media perception; (5) large-scale perception information processing and learning engine; (6) urban full-dimension intelligent perception reasoning engine. These research directions should be focused on the future AI 2.0 research plan.