Automatic recognition of insect sounds using MFCC and GMM

昆虫的运动、取食、鸣叫都会发出声音,这些声音存在种内相似性和种间差异性,因此可用来识别昆虫的种类.基于昆虫声音的昆虫种类自动检测技术对协助农业和林业从业人员方便地识别昆虫种类非常有意义.本研究采用了语音识别领域里的声音参数化技术来实现昆虫的声音自动鉴别.声音样本经预处理后,提取梅尔倒谱系数(Mel-frequency cepstrum coefficient,MFCC)作为特征,并用这些样本提取的MFCC特征集训练混合高斯模型(Gaussian mixture model,GMM).最后用训练所得到的GMM对未知类别的昆虫声音样本进行分类.该方法在包含58种昆虫声音的样本库中进行了评估,取得了较高的识别正确率(平均精度为98.95％)和较理想的时间性能.该测试结果证明了基于MFCC和GMM的语音参数化技术可以用来有效地识别昆虫种类.     

基于Mel倒谱系数和矢量量化的昆虫声音自动鉴别

为了给生产单位害虫管理的普通技术人员提供简便易操作的昆虫种类鉴别方法, 本研究把人类语音识别领域的先进技术应用于昆虫识别, 提出了一种新颖的昆虫声音自动鉴别方法, 用声音参数化技术为昆虫声纹识别设计了一种简单易行的方案。声音信号经过预处理、分段得到一系列的声音样本, 从声音样本提取Mel倒谱系数（MFCC）, 并用Linde-Buzo-Gray（LBG）算法对提取的MFCC进行矢量量化(VQ), 所得码字作为声音样本的特征模型。特征参数之间的匹配用搜索最近邻的方法实现。本文方法在包含70种昆虫声音的库中进行了试验, 取得了超过96%的识别率和理想的时间性能。试验结果证明了该方法的有效性。     

基于颜色名和OpponentSIFT特征的鳞翅目昆虫图像识别

【目的】本研究旨在探索使用先进的计算机视觉技术实现对昆虫图像的自动分类方法。【方法】通过预处理对采集的昆虫标本图像去除背景,获得昆虫图像的前景蒙板,并由蒙板确定的轮廓计算出前景图像的最小包围盒,剪切出由最小包围盒确定的前景有效区域,然后对剪切得到的图像进行特征提取。首先提取颜色名特征,把原来的RGB（Red-Green-Blue）图像的像素值映射到11种颜色名空间,其值表示RGB值属于该颜色名的概率,每个颜色名平面划分成3×3像素大小的网格,用每格的概率均值作为网格中心点的描述子,最后用空阈金字塔直方图统计的方式形成颜色名视觉词袋特征;其次提取OpponentSIFT(Opponent Scale Invariant Feature Transform)特征,首先把RGB图像变换到对立色空间,对该空间每通道提取SIFT特征,最后用空域池化和直方图统计方法形成OpponentSIFT视觉词袋。将两种词袋特征串接后得到该昆虫图像的特征向量。使用昆虫图像样本训练集提取到的特征向量训练SVM（Support Vector Machine）分类器,使用这些训练得到的分类器即可实现对鳞翅目昆虫的分类识别。【结果】该方法在包含10种576个样本的昆虫图像数据库中进行了测试,取得了100%的识别正确率。【结论】试验结果证明基于颜色名和OpponentSIFT特征可以有效实现对鳞翅目昆虫图像的识别。     

迁飞昆虫生物学参数反演及种类辨识分析

虫害严重威胁着我国的粮食安全,迁飞昆虫中有许多是农业害虫,其远距离迁飞是导致虫害异地暴发的重要原因.昆虫雷达是观测昆虫迁飞最有效的工具,在迁飞昆虫的监测和预警中发挥着越来越重要的作用,但传统昆虫雷达不能准确获取昆虫的各项生物学参数,因此无法实现昆虫种类的精确识别.随着雷达技术的创新和发展,通过昆虫雷达获得较为准确的昆虫生物学参数成为可能,为基于昆虫雷达实现迁飞昆虫个体种类辨识提供了依据.本文综述了从雷达回波中提取多频和极化散射参量,然后基于不同的电磁散射反演昆虫生物学参数的方法,并对比分析了基于不同方法的昆虫体重、体长、体宽和振翅频率的反演精度.最后基于生物学参数,采用5种机器学习算法以高精度实现了 23种迁飞昆虫的种类辨识,并分析了昆虫生物学参数的测量误差对迁飞昆虫种类辨识精度的影响,初步验证了利用雷达实现高精度迁飞昆虫种类辨识的可行性.     

基于稀疏编码和SCG BPNN的鳞翅目昆虫图像识别

【目的】为了给林业、 农业或植物检疫等行业人员提供一种方便快捷的昆虫种类识别方法, 本文提出了一种新颖的鳞翅目昆虫图像自动识别方法。【方法】首先通过预处理对采集的昆虫标本图像去除背景, 分割出双翅, 并对翅图像的位置进行校正。然后把校正后的翅面分割成多个超像素, 用每个超像素的l, a, b颜色及x, y坐标平均值作为其特征数据。接下来用稀疏编码(SC)算法训练码本、 生成编码并汇集成特征向量训练量化共轭梯度反向传播神经网络（SCG BPNN）, 并用得到的BPNN进行分类识别。【结果】该方法对包含576个样本的昆虫图像的数据库进行了测试, 取得了高于99%的识别正确率, 并有理想的时间性能、 鲁棒性及稳定性。【结论】实验结果证明了本文方法在识别鳞翅目昆虫图像上的有效性。     

实蝇科果实蝇属昆虫数字图像自动识别系统的构建和测试

针对双翅目实蝇科果实蝇属昆虫的自动识别,本文提出利用翅及中胸背板图像的局部二进制模式（local binary pattern, LBP）特征,采用Adaboost算法, 设计和开发“实蝇科果实蝇属昆虫数字图像自动识别系统”（Automated Fruit fly Identification System-Bactrocera, AFIS-B）。该系统包括图像采集、图像裁剪、预处理、特征提取、分类器设计、识别和显示,共7个模块。研究结果表明： LBP特征可以有效鉴别实蝇科果实蝇属昆虫;在对实蝇科果实蝇属8个种的测试中, 该系统表现出较高的准确性和稳定性,平均识别率可达80%以上。此外,还对果实蝇属昆虫翅膀及中胸背板图像在光照不均匀、姿态扭曲、样本受损及样本量大小等不同条件下的识别率进行了试验测试。结果表明, 该系统对测试样本的光照不均匀、 姿态扭曲和样本受损都表现出良好的鲁棒性,正确识别率与训练集样本各个种数量在一定条件下明显正相关,与训练集样本物种总量负相关。该项研究为实蝇科有害昆虫自动识别系统的构建及实际应用提供了理论、 方法及基础数据的支撑, 亦可为其他昆虫自动识别系统的研究和构建提供有益借鉴。 关键词：     

试论昆虫种类鉴定的准确性

昆虫分类学有３个主要的任务：（１）鉴定昆虫种类;（２）研究昆虫的区系;（３）探索昆虫的自然系统发育。这３项工作既相互联系,又有各自的侧重点。鉴定昆虫种类的核心问题是识别昆虫是哪一个种,它的正确学名（有效名）是什么。在许多类似的昆虫中区别它们是否为同一个种。１物种概念是昆虫鉴定的指导思想物种是什么？依据什么来确定一个物种？这是昆虫鉴定最基本的问题。这一概念的理解对鉴定工作有着最直接的影响。昆虫种类鉴定就是在这一概念指导下进行的。物种的定义有很多,花保祯博士曾在其博士论文中例举了最具代表性的一些定义…     

四种网蝽科昆虫外部形态特征提取与分析

随着计算机图像处理技术的发展,昆虫图像特征提取技术在昆虫分类学中的应用越来越广泛。相关的研究报道涉及到多种较大型昆虫和节肢动物,但对身体较小昆虫的外部形态特征提取与分析,及其雌雄鉴别方面的报道不多。本研究利用中国农业大学IPMist实验室开发的Bug Shape 1.0对体型较小的4种网蝽科Tingidae昆虫成虫进行了外部形态特征的提取,分析了所得参数在其种类和雌雄方面的差异。结果表明,获得的10个参数在4种网蝽成虫中差异均有高度统计学意义。4种网蝽雌雄成虫在面积、等效圆半径、偏心率、球状性和圆形度5个参数方面的差异有高度统计学意义,初步认为可以作为区别这4种网蝽雌雄成虫的重要指标。判别分析证明,4种网蝽交叉判别的正确率高达98.5%,对种类和性别同时进行判别时,交叉判别的正确率为85.0%。利用重心法得到的聚类分析结果符合传统昆虫分类学的结果,同一属的菊方翅网蝽Corythucha marmorata和悬铃木方翅网蝽C.ciliata优先聚为一类,其他2种网蝽先后分别聚类。本研究利用图像处理与分析技术得到的参数适合对此类昆虫进行种类和性别的鉴定,初步表明该技术在此类害虫分类鉴定中的科学性和可行性。     

DNA条形码技术在北京百花山地区夜蛾科物种鉴定中的应用

为了探讨DNA条形码技术在夜蛾物种鉴定中的可行性, 本研究利用条形码通用引物扩增了北京百花山地区43种夜蛾75个样本的线粒体细胞色素C氧化酶亚基I （mitochondrial cytochrome c oxidase subunit I, COI）基因序列, 以Kimura双参数模型进行种内种间遗传距离分析、 使用邻接法（neighbor-joining, NJ）和最大简约法（maximum parsimony, MP）分别构建系统发育树, 并利用分子序列差异阈值对样本进行分子可操作分类单元（molecular defined operational taxonomic units, MOTU）划分。结果表明： 所有夜蛾种类通过系统发育树可以成功区分; 种内平均遗传距离(0.03%)远远小于种间平均遗传距离(11.29%); 采用较为保守的1%的序列差异阈值将75个夜蛾样本分为42个MOTU, 正确率为95%, 除了MOTU04包含2个物种外, 剩余41个MOTU与形态种呈现一一对应的关系。结果显示, 基于COI基因的DNA条形码对于本研究中所涉及的夜蛾具有较好的区分, 可以作为一种有效的工具在夜蛾科昆虫物种鉴定中进行应用。     

Roles of Hydrocarbons in the Recognition Systems of Insects

SYNOPSIS. Many bioassays have shown that cuticular hydrocarbonsare used in the recognition systems of both solitary and socialinsects. The function of insect recognition systems is to enablean insect to recognize, and possibly discriminate, its own species,sex, or kin from that of other insects. The primary functionof cuticular hydrocarbons is to protect the insects from desiccation.Hydrocarbons can be removed from insect cuticles and characterizedwith gas chromatography/mass spectrometry. Studies using suchanalytical techniques have revealed that insect hydrocarboncompositions are species-specific, sex-specific and, in socialinsects, colony- and caste-specific. Furthermore, recognitionbioassays have confirmed that certain components of the cuticleof some insect species are sex attractants as well as aphrodisiacsor sex inhibitors. Other bioassays have shown that hydrocarbonsare important in facilitating colony structure in social insects.In addition, the hydrocarbons of some parasitic insects appearto mimic those of their host species. Thus, hydrocarbons areproving to be very important in the everyday activities of manyinsect species.     

Recognition of Non-Harmonic Natural Sounds by Small Mammals Using Competitive Training

Animals recognize biologically relevant sounds, such as the non-harmonic sounds made by some predators, and respond with adaptive behaviors, such as escaping. To clarify which acoustic parameters are used for identifying non-harmonic, noise-like, broadband sounds, guinea pigs were conditioned to a natural target sound by introducing a novel training procedure in which 2 or 3 guinea pigs in a group competed for food. A set of distinct behavioral reactions was reliably induced almost exclusively to the target sound in a 2-week operant training. When fully conditioned, individual animals were separately tested for recognition of a set of target-like sounds that had been modified from the target sound, with spectral ranges eliminated or with fine or coarse temporal structures altered. The results show that guinea pigs are able to identify the noise-like non-harmonic natural sounds by relying on gross spectral compositions and/or fine temporal structures, just as birds are thought to do in the recognition of harmonic birdsongs. These findings are discussed with regard to similarities and dissimilarities to harmonic sound recognition. The results suggest that similar but not identical processing that requires different time scales might be used to recognize harmonic and non-harmonic sounds, at least in small mammals.     

Classification of insects by echolocating greater horseshoe bats

Summary Echolocating greater horseshoe bats (Rhinolophus ferrumequinum) detect insects by concentrating on the characteristic amplitude- and frequency modulation pattern fluttering insects impose on the returning echoes. This study shows that horseshoe bats can also further analyse insect echoes and thus recognize and categorize the kind of insect they are echolocating.Four greater horseshoe bats were trained in a twoalternative forced-choice procedure to choose the echo of one particular insect species turning its side towards the bat (Fig. 1). The bats were able to discriminate with over 90% correct choices between the reward-positive echo and the echoes of other insect species all fluttering with exactly the same wingbeat rate (Fig. 4).When the angular orientation of the reward-positive insect was changed (Fig. 2), the bats still preferred these unknown echoes over echoes from other insect species (Fig. 5) without any further training. Because the untrained bats did not show any prey preference, this indicates that the bats were able to perform an aspect-anglein-dependent classification of insects.Finally we tested what parameters in the echo were responsible for species recognition. It turned out that the bats especially used the small echo-modulations in between glints as a source of information (Fig. 7). Neither the amplitudenor the frequencymodulation of the echoes alone was sufficient for recognition of the insect species (Fig. 8). Bats performed a pattern recognition task based on complex computations of several acoustic parameters, an ability which might be termed cognitive.Abbreviations AM amplitude modulation - CF constant frequency - FM frequency modulation - S+ positive stimulus - S- negative stimulus     

A PCR method for detection of plant meals from the guts of insects

The feeding behaviour of insects is a difficult ecological interaction to study. To date, entomologists have used biochemical and molecular techniques to identify the meals of predatory insects. We present here a molecular approach to identifying the DNA of plant species in the insect gut using the ribulose bisphosphate carboxylase gene large subunit (rbcL). A reference collection of 23 plant species from the southern Jordan Valley, Israel, was genetically characterized and employed. Insects belonging to eight different families were collected in the field along with the plants upon which they were found. After collection and prior to analysis, these insects were isolated on the plants they were found upon in the laboratory. This was to ensure that the insects had only one plant meal in their gut, as multiple plant meals would require additional techniques like cloning. A blind study was performed, genetically confirming plant DNA to species level from the processed gut contents of the insects. All reference plant species could be differentiated using a 157 bp long fragment of the rbcL gene. Plant DNA was identifiable, and the meal of the respective insect was accurately determined in each case. Analyses using experimentally fed crickets, Gryllodes hebraeus, determined that plant DNA was still detectable by PCR up to 12 h post-ingestion. This research proposes the application of molecular techniques for the identification of herbivorous insect feeding behaviour to increase understanding of plant–insect interactions.     

Aspects of Species Recognition by Sound in Four Species of Damselfishes,Genus Eupomacentrus (Pisces: Pomacentridae)

Members of four sympatric species of Eupomacentrus carry out reproductive activities at the same time of the year and produce similar pulsed courtship sounds. Such sounds are known to facilitate courtship among conspecifics. Consequently, members of the four species in the field and in the laboratory were tested with the various sounds to determine if they could distinguish their own species sounds from those produced by congeners. The differential responses clearly demonstrate species specific recognition by sound and indicate that the pulse interval and the number of pulses per sound are the important parameters for this recognition.     

Kin-informative recognition cues in ants

Although social groups are characterized by cooperation, they are also often the scene of conflict. In non-clonal systems, the reproductive interests of group members will differ and individuals may benefit by exploiting the cooperative efforts of other group members. However, such selfish behaviour is thought to be rare in one of the classic examples of cooperation--social insect colonies--because the colony-level costs of individual selfishness select against cues that would allow workers to recognize their closest relatives. In accord with this, previous studies of wasps and ants have found little or no kin information in recognition cues. Here, we test the hypothesis that social insects do not have kin-informative recognition cues by investigating the recognition cues and relatedness of workers from four colonies of the ant Acromyrmex octospinosus. Contrary to the theoretical prediction, we show that the cuticular hydrocarbons of ant workers in all four colonies are informative enough to allow full-sisters to be distinguished from half-sisters with a high accuracy. These results contradict the hypothesis of non-heritable recognition cues and suggest that there is more potential for within-colony conflicts in genetically diverse societies than previously thought.     

Geophone detection of subterranean termite and ant activity

A geophone system was used to monitor activity of subterranean termites and ants in a desert environment with low vibration noise. Examples of geophone signals were recorded from a colony of Rhytidoponera taurus (Forel), a colony of Camponotus denticulatus Kirby, and a termite colony (undetermined Drepanotermes sp.) under attack by ants from a nearby C. denticulatus colony. The geophone recordings were compared with signals recorded from accelerometers in a citrus grove containing Solenopsis invicta Buren workers. Because of their small size, all of these insects produce relatively weak sounds. Several different types of insect-generated sounds were identified in the geophone recordings, including high-frequency ticks produced by R. taurus and C. denticulatus, and patterned bursts of head bangs produced by Drepanotermes. The S. invicta produced bursts of ticks with three different stridulation frequencies, possibly produced by three different-sized workers. Overall, both systems performed well in enabling identification of high-frequency or patterned pulses. The geophone was more sensitive than the accelerometer to low-frequency signals, but low-frequency insect sound pulses are more difficult to distinguish from background noises than high-frequency pulses. The low cost of multiple-geophone systems may facilitate development of future applications for wide-area subterranean insect monitoring in quiet environments.     

If a bird flies in the forest,does an insect hear it?

Birds are major predators of many eared insects including moths, butterflies, crickets and cicadas. We provide evidence supporting the hypothesis that insect ears can function as ‘bird detectors’. First, we show that birds produce flight sounds while foraging. Eastern phoebes (Sayornis phoebe) and chickadees (Poecile atricapillus) generate broadband sounds composed of distinct repetitive elements (approx. 18 and 20 Hz, respectively) that correspond to cyclic wing beating. We estimate that insects can detect an approaching bird from distances of at least 2.5 m, based on insect hearing thresholds and sound level measurements of bird flight. Second, we show that insects with both high and low frequency hearing can hear bird flight sounds. Auditory nerve cells of noctuid moths (Trichoplusia ni) and nymphalid butterflies (Morpho peleides) responded in a bursting pattern to playbacks of an attacking bird. This is the first study to demonstrate that foraging birds generate flight sound cues that are detectable by eared insects. Whether insects exploit these sound cues, and alternatively, if birds have evolved sound-reducing foraging tactics to render them acoustically ‘cryptic’ to their prey, are tantalizing questions worthy of further investigation.