发声生物体的行为智能分析与仿生声纳技术研究

Through millions of years’ evolution, the mammalian ancestors of bats and cetaceans, they have developed a sophisticate and precise acoustical detection, positioning, navigation and communication system. They can adjust project signal, detection logistic adaptive to the various environment and prey target, for the purpose of biological function. However, these kinds of nature intelligence were neglected since sonar designer thought through artificial ways. Where the deepen understanding to observation and analysis, biomimetic sonar will unveil step by step. This proposal aims to study the common advantage among nature biological acoustic source, explore the intelligence behinds signal, and verify the biomimetic sonar through achievable way. Based on published results, multi-scale and multi-harmonic characters extract from biological acoustic pulse signals, especially for the purpose of forge. Logics for the sensing and detection will be extract via environmental reconstruction and sound field simulation. Such as low duty cycle echolocators cannot broadcast and receive at the same time and this limits the duration of biosonar signals throughout the process of finding and catching a prey item. Furthermore, echo signal with two-way- travel time through the environment plus potential appeared target is processed in the brain. In biomimetic sonar, the adaptive network-based fuzzy inference with echo signal multiple features exaction will be most similar way as real nature animal. And specials conjugation pair signal for the purpose to classify target from bubbles will be another great biological intelligence of wild dolphin. All the technique will be optimized through independent, random events scenario generator simulation. Finally, the field experiment will be developed, just like the real environment for piloting, searching and detection suffering different conditions such as high interferences of ship radiated noise, schools, nets, bubble cloud. All these efforts should be devote to improving the intelligence of sonar, especially for AUVs.

经过自然界的进化，鲸豚等水中哺乳动物及空气中的蝙蝠类生物，发展了完善的声学探测、定位、导航、通信系统，它们能够根据周边的环境、捕食对象变化而适应性调整发射信号、探测逻辑，达到其功能性目的。而目前的人工声纳还没有这样的智能，如何从生物信号仿生角度分析和探索其智能优势，并通过可实践的方法对仿生声纳性能进行检验是本项目研究的主要目标。本项目将对发声生物信号特征的共同特征进行分析，分析其多谐波、联合尺度的共同特征。在环境重构和声场建模基础上，分析生物体探测对周边环境的感知和目标探测过程中的探测逻辑。进一步尝试探索生物体利用信号的特征联合提取、环境适应的多通道数据融合目标判决智能，分析仿生共轭脉冲对信号的目标识别优势。通过独立随机事件群效能评估仿真和目标探测识别的仿生声纳检验试验对上述智能行为进行检验。为仿生技术推动声纳智能化，尤其是AUV等无人平台的声纳发展提供支持。

围绕水下如何理解并有效利用鲸豚、蝙蝠等发声动物生物智能的前沿性问题，本项目研究突破模拟生物信号波形“形似”局限，开展了生物发声声学信号的波形结构、探测逻辑与环境态势、使命任务之间的耦合关联、基于模型与数据联合驱动的场景重构的生物发声行为智能解构和海上外场无人平台载荷探测试验验证。从生物信号行为分析、智能提取到智能重现三个层面提升对了生物发声探测感知行为逻辑的认知，从而为发声生物自然界物竞天择进化获得的智能行为“神似”的人工仿生声纳应用奠定了技术基础。.在课题研究期间本项目具体研究内容突破如下：.1..发声生物信号及发声逻辑的认知研究.充分借鉴和吸收现有国内外对于生物发声现象的研究成果，总结发声哺乳动物的发声信号的类型、特征，梳理了信号波形、信号参数、发声间隔等发声逻辑的自适应变化规律，并应用与仿生探测声呐参数调节，为试验验证仿生智能提供了有力抓手。.2..发声生物信号探测逻辑场景重构解析.通过模型和数据双重驱动，重构了探测场景，在探测任务约束下模拟生物所用声信号在时频结构上的特点，以及环境适应、目标适应的信号参数调节机制，设计了仿生声纳探测波形调节机制。利用探测场景重构、声场重构，解析验证了生物智能的探测优势。.3..仿生多谐波共轭探测算法.设计了一种多谐波仿生共轭脉冲对探测波形及多尺度匹配处理方法相结合的探测方案，该方法能有效抑制稳定的混响的同时，突出移动目标回波。水池试验结果验证了该方法的可行性及优越性。.4..海上平台仿生探测能力验证.在威海海域开展了相关仿生探测验证试验，测试了仿生声纳探测波形调节机制的有效性。. 在项目支持下，共计发表论文11篇，其中SCI论文9篇，第一标注SCI论文8篇，第二标注SCI论文1篇。申请发明专利5项，其中授权1项。培养毕业硕士研究生5名，博士研究生2名。相关成果在极地复杂环境中的仿虎鲸信号声学测量、水下探测国防类重大项目中得到应用。