NL2029582B1 - Method for intelligently identifying estrus state through sheep sound - Google Patents
Method for intelligently identifying estrus state through sheep sound Download PDFInfo
- Publication number
- NL2029582B1 NL2029582B1 NL2029582A NL2029582A NL2029582B1 NL 2029582 B1 NL2029582 B1 NL 2029582B1 NL 2029582 A NL2029582 A NL 2029582A NL 2029582 A NL2029582 A NL 2029582A NL 2029582 B1 NL2029582 B1 NL 2029582B1
- Authority
- NL
- Netherlands
- Prior art keywords
- sheep
- sound
- estrus
- state
- intelligently identifying
- Prior art date
Links
Classifications
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01K—ANIMAL HUSBANDRY; CARE OF BIRDS, FISHES, INSECTS; FISHING; REARING OR BREEDING ANIMALS, NOT OTHERWISE PROVIDED FOR; NEW BREEDS OF ANIMALS
- A01K29/00—Other apparatus for animal husbandry
- A01K29/005—Monitoring or measuring activity, e.g. detecting heat or mating
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Environmental Sciences (AREA)
- Biophysics (AREA)
- Animal Husbandry (AREA)
- Biodiversity & Conservation Biology (AREA)
- Auxiliary Devices For Music (AREA)
Abstract
The invention discloses a method for intelligently identifying estrus state through sheep sound, comprising the following steps: Sl, collecting sheep sound; S2, cutting and pre-processing the sheep sound to obtain estrus sound segment; S3, comparing frequency 5 domain analysis and energy on the estrus sound segment to obtain estrus characteristic parameters; S4, constructing a sheep sound recognition model based on the parameters of S3, obtaining sheep estrus state result based on the sheep sound recognition model. In the invention, useless components can be filtered out by cutting sheep's calls, which is convenient for later model training, and the accuracy of sound analysis and recognition is 10 improved by pre-treatment, so as to obtain a high-precision sound and audio domain, thereby completing the construction of sheep's sound recognition model, obtaining recognition results and analyzing sheep's calls, avoiding unnecessary loss of manpower and financial resources, improving identification- monitoring efficiency, and realizing animals welfare breeding.
Description
Method for intelligently identifying estrus state through sheep sound
The invention relates to the technical field of acoustic analysis, in particular to a method for intelligently identifying estrus state through sheep sound.
Animal husbandry has entered a new stage of development, and is transforming from traditional animal husbandry to modern animal husbandry, so as to build resource-saving and environment-friendly animal husbandry. Traditional animal husbandry mainly relies on manual observation to monitor animal condition, which is not only inefficient but also easy to interfere with the living environment of cattle and sheep. With the development of fine animal husbandry, it has become an important trend to monitor and analyze animal behavior, health and welfare with intelligence and information. In the process of breeding herbivores such as cattle and sheep, because herbivores generally have a wide range of activities, and there is often a breeding mode combining stocking and captivity, it is difficult to monitor various behaviors during grazing.
Under the current environment, sheep still have low efficiency, great waste of labor cost and time cost in the fields of intensive breeding, production and reproduction, so there is great room for development in the field of intelligent identification. However, the aquaculture industry is developing from extensive free-range breeding to large-scale intensive centralized breeding, and the intensive degree of sheep is still relatively backward. For the identification of estrus, only the traditional test method can be used to judge estrus or not, which cannot be intelligently identified according to the sound of sheep.
The instantaneity is not strong, and it is easy to produce mismatches and missing matches, and the real-time monitoring methods are limited.
In view of the above problems, the invention provides a method for intelligently identifying estrus state through the sound of sheep, so as to solve the technical problems existing in the prior art. By analyzing the calls of sheep, unnecessary loss of manpower and financial resources can be avoided, and the efficiency of identification and monitoring can be improved, and at the same time, welfare breeding of animals can be realized.
In order to achieve the above purpose, the invention provides a method for intelligently identifying estrus state by sheep's sound, which comprises the following steps:
S1, collecting the sound of sheep;
S2, cutting and pre-processing the sheep sound to obtain an estrus sound segment;
S3, performing frequency domain analysis and energy comparison on the estrus sound segment to obtain estrus characteristic parameters;
S4, constructing a sheep sound recognition model based on the estrus characteristic parameters, and obtaining a sheep estrus state result based on the sheep sound recognition model.
Preferably, the sheep's sound in S1 includes ewe estrus, ewe hunger, ewe being stimulated by feed, and lamb crying for mother.
Preferably, the acquisition method in S1 includes simulating the scene environment for natural recording.
Preferably, the cutting process in S2 includes cutting the collected sheep's sound into several audio signals, deleting the noise in the audio signals, and obtaining the sheep's audio signals.
Preferably, the pre-processing in S2 includes sound denoising, endpoint analysis and windowing and framing.
Preferably, the sound denoising includes mixing Gaussian noise into the audio signal of the sheep, and obtaining the denoising audio by using a wavelet threshold method.
Preferably, the endpoint analysis includes performing double-threshold endpoint detection on the denoising audio to obtain zero-crossing rate and energy, and obtaining the estrus audio position according to the zero-crossing rate and energy.
Preferably, the windowing and framing comprises: framing and windowing the denoising audio according to the position of the estrus audio, and obtaining an estrus sound segment.
The invention disclose that following technical effect:
According to the method, useless components can be filtered out by cutting sheep calls, which is convenient for later model training; and the accuracy of sound analysis and recognition is improved by pretreatment, so as to obtain a high-precision sound and audio domain, thereby completing the construction of a sheep sound recognition model, obtaining recognition results and analyzing sheep calls, avoiding unnecessary loss of manpower and financial resources, improving the efficiency of identification and monitoring, and realizing welfare breeding of animals.
In order to explain the embodiments of the invention or the technical scheme in the prior art more clearly, the following will briefly introduce the drawings used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the invention, and for ordinary technicians in the field, other drawings can be obtained according to these drawings without paying creative labor.
Fig. 1 is a flow chart of the method of the invention;
Fig. 2 is a schematic diagram of sound recognition in an embodiment of the invention.
The following will clearly and completely describe the technical scheme in the embodiments of the invention with reference to the drawings in the embodiments of the invention. Obviously, the described embodiments are only part of the embodiments of the invention, not all of them. Based on the embodiments of the invention, all other embodiments obtained by ordinary technicians in the field without creative labor belong to the scope of protection of the invention.
In order to make the above objects, features and advantages of the invention more obvious and easy to understand, the invention will be further explained in detail with reference to the drawings and specific embodiments.
Referring to fig. 1-2, this embodiment provides a method for intelligently identifying estrus state by sheep's sound, which includes the following steps:
S1, collecting the sound of sheep, in order to simulate the field environment, 40 multiparous ewes are recorded in-situ, and the calls of ewes in estrus, ewes hungry, ewes stimulated by feed and lambs seeking mothers are recorded. As sound samples, the sound samples include training samples and recognition samples.
S2, cutting and pre-processing sheep sound to obtain estrus sound segments.
Sheep sound signal cutting processing: cutting continuous sound samples into separate sounds and deleting useless noise components. In each state, 32 separate audio signals are randomly used, and a total of 128 sounds are used for model training.
Pretreatment of sheep sound: Pretreatment of sheep sound through three steps. The first step is sound denoising: firstly, Gaussian noises with other signal-to-noise ratios are mixed into the audio signals of sheep, and the noisy signals of sheep's barking sounds are simulated. Then, the wavelet threshold method is used to denoise the sounds, so as to obtain the denoised frequency and avoid insufficient recognition and analysis accuracy;
Secondly, endpoint detection and analysis: using double-threshold endpoint detection method, according to the zero-crossing rate and energy of the denoised audio signal, the position of the effective sound in a whole section of sound is determined, and the sheep sound signal of each frame is subjected to short-term average zero-crossing rate operation.
Mark the starting point and ending point of a sound. The third step is windowing and framing: using the method of framing and windowing to cut out the sound segments with relatively stable spectral characteristics, obtain the position of estrus audio, and divide the denoised audio signal by window function, taking "frame" as the unit.
Wherein, the wavelet threshold method is as follows: (1) carrying out wavelet transform on a noisy signal x(t) to obtain a group of wavelet decomposition coefficients Wijk; ; (2) thresholding wavelet decomposition coefficients to obtain wavelet estimation coefficients which are as small as possible; (3) The estimated wavelet coefficients are used for wavelet reconstruction to obtain an estimated signal, which is the de-noised signal.
The process of double-threshold endpoint detection method is as follows: (1) before starting endpoint detection, two thresholds are respectively determined for short-time energy and zero-crossing rate; (2) Set a relatively low threshold, which is relatively small and sensitive to signal changes; (3) Set a relatively high threshold, whose value is relatively large, and the signal must reach a certain intensity; (4) When the low threshold is exceeded, it is not necessarily the beginning of speech,
but it may be caused by short-time noise. If the high threshold is exceeded, it can be basically believed that it is caused by the sound signal.
S3, performing frequency domain analysis and energy comparison on the estrus sound segment to obtain estrus characteristic parameters. 5 Time-domain and frequency-domain analysis of sheep sound signal: The differences of four kinds of sounds are compared through time-domain and frequency analysis. The frequency domain diagram can be obtained by Fourier transform in the time domain.
Energy comparison of sheep sound signal: the recording pen with frequency of 48 000Hz can be used for sampling, and energy analysis can be carried out under the condition of the same distance and pitch to obtain the characteristic parameters of estrus.
S4, constructing a sheep sound recognition model based on the estrus characteristic parameters, and obtaining a sheep estrus state result based on the sheep sound recognition model.
The sheep sound recognition model adopts HMM model. Firstly, the HMM model is initialized, and the characteristic parameters randomly extracted from the sheep sound samples are taken as observation sequences. Then, the HMM model obtained by training is calculated by using the bidirectional calculation algorithm. Input the sheep sound signal to be recognized into the trained model, determine whether the detection sequence corresponds to the model parameters, and finally recognize and output the results.
In this embodiment, on the premise of no noise interference, after the induction of the ram, the sound of each stage (pre-estrus, estrus, post-estrus and estrus) of the ewe in estrus is recorded, and then the sound of the normal ewe is recorded, and the pretreatment such as shearing, denoising and endpoint detection is performed to obtain the characteristic parameters of estrus.
The acquisition of the characteristic parameters of the model includes: the state is defined as four states: estrus, sheep seeking mother, hunger and feed stimulation, so the probability after initialization is” =[L0.0, 0] and the probability of the four states in the initial state is as follows:
0.5 05 0 0 u 0 05050
P= 0 0505 0 0 0 1
Randomly extract a segment of sheep's sound in four states to extract characteristic parameters, and express the observation sequence as © (5, %,, % +.) Among them, i is the type of cry, n is the kind of cry. The characteristic parameters extracted from random samples are taken as observation sequences, and the conditional probabilities of thresholds in the trained HMM model are calculated by using the bidirectional calculation algorithm.
Re-take the ewes' calls (the state 1s random, but the state must be recorded), divide them into separate calls, and automatically recognize them through the trained program.
Write the main interface program of input/output, analyze the ewe's sound directly in frequency domain and print the results on the main interface.
The invention disclose that following technical effect:
According to the method, useless components can be filtered out by cutting sheep calls, which is convenient for later model training; and the accuracy of sound analysis and recognition 1s improved by pretreatment, so as to obtain a high-precision sound and audio domain, thereby completing the construction of a sheep sound recognition model, obtaining recognition results and analyzing sheep calls, avoiding unnecessary loss of manpower and financial resources, improving the efficiency of identification and monitoring, and simultaneously realizing welfare breeding of animals.
Finally, it should be noted that the above-mentioned embodiments are only concrete embodiments of the invention, and are used to illustrate the technical scheme of the invention, but not to limit it. Although the invention has been described in detail with reference to the above-mentioned embodiments, ordinary people in the field should understand that any person familiar with the technical field can still modify or easily think of changes to the technical scheme described in the above-mentioned embodiments within the technical scope disclosed by the invention, or replace some of the technical features equally. However, these modifications, changes or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be covered within the protection scope of the present invention. Therefore, the protection scope of the invention shall be subject to the protection scope of the embodiments.
PREFERRED EMBODIMENTS OF THE INVENTION ARE AS FOLLOWS: 1. A method for intelligently identifying estrus state by sheep's sound, which is characterized by comprising the following steps:
S1, collecting the sound of sheep;
S2, cutting and pre-processing the sheep sound to obtain an estrus sound segment;
S3, performing frequency domain analysis and energy comparison on the estrus sound segment to obtain estrus characteristic parameters;
S4, constructing a sheep sound recognition model based on the estrus characteristic parameters, and obtaining a sheep estrus state result based on the sheep sound recognition model. 2. The method for intelligently identifying estrus state by sheep's sound according to embodiment 1, characterized in that the sheep's sound in S1 includes ewe's estrus, ewe's hunger, ewe's being stimulated by feed and lamb's cry of seeking mother. 3. The method for intelligently identifying estrus state by sheep's sound according to embodiment], characterized in that the collection method in S1 includes simulating the scene environment for natural recording. 4. The method for intelligently identifying estrus state by sheep's sound according to embodiment 1, characterized in that the cutting process in S2 comprises cutting the collected sheep's sound into several audio signals, deleting the noise in the audio signals, and obtaining the sheep's audio signals. 5. The method for intelligently identifying estrus state by sheep's sound according to embodiment 4, characterized in that the pre-processing in S2 includes sound denoising, endpoint analysis and windowing and framing. 6. The method for intelligently identifying estrus state by sheep's sound according to embodiment 5, characterized in that the sound denoising comprises mixing Gaussian noise into the audio signal of the sheep, and obtaining the denoising audio by using a wavelet threshold method. 7. The method for intelligently identifying estrus state by sheep's sound according to embodiment 6, characterized in that the endpoint analysis comprises performing double-threshold endpoint detection on the denoising audio to obtain zero-crossing rate and energy, and obtaining the estrus audio position according to the zero-crossing rate and energy. 8. The method for intelligently identifying estrus state by sheep's sound according to embodiment 7, characterized in that the windowing and framing comprises framing and windowing the denoising audio according to the estrus audio position to obtain the estrus sound segment.
Claims (8)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
NL2029582A NL2029582B1 (en) | 2021-11-02 | 2021-11-02 | Method for intelligently identifying estrus state through sheep sound |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
NL2029582A NL2029582B1 (en) | 2021-11-02 | 2021-11-02 | Method for intelligently identifying estrus state through sheep sound |
Publications (1)
Publication Number | Publication Date |
---|---|
NL2029582B1 true NL2029582B1 (en) | 2023-06-01 |
Family
ID=82781094
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
NL2029582A NL2029582B1 (en) | 2021-11-02 | 2021-11-02 | Method for intelligently identifying estrus state through sheep sound |
Country Status (1)
Country | Link |
---|---|
NL (1) | NL2029582B1 (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106719066A (en) * | 2017-03-27 | 2017-05-31 | 南京农业大学 | A kind of bionics ewe is looked into feelings/lure feelings device and looks into feelings/lure feelings method |
CN106847293A (en) * | 2017-01-19 | 2017-06-13 | 内蒙古农业大学 | Facility cultivation sheep stress behavior acoustical signal monitoring method |
CN111467074A (en) * | 2020-05-18 | 2020-07-31 | 北京海益同展信息科技有限公司 | Method and device for detecting the state of animals |
-
2021
- 2021-11-02 NL NL2029582A patent/NL2029582B1/en active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106847293A (en) * | 2017-01-19 | 2017-06-13 | 内蒙古农业大学 | Facility cultivation sheep stress behavior acoustical signal monitoring method |
CN106719066A (en) * | 2017-03-27 | 2017-05-31 | 南京农业大学 | A kind of bionics ewe is looked into feelings/lure feelings device and looks into feelings/lure feelings method |
CN111467074A (en) * | 2020-05-18 | 2020-07-31 | 北京海益同展信息科技有限公司 | Method and device for detecting the state of animals |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jiang et al. | Whistle detection and classification for whales based on convolutional neural networks | |
CN109258509B (en) | Intelligent monitoring system and method for abnormal sound of live pigs | |
CN109243470A (en) | Broiler chicken cough monitoring method based on Audiotechnica | |
CN113539294A (en) | Method for collecting and identifying sound of abnormal state of live pig | |
CN110189756A (en) | It is a kind of for monitoring the method and system of live pig abnormal sound | |
CN112820275A (en) | Automatic monitoring method for analyzing abnormality of suckling piglets based on sound signals | |
Henríquez et al. | An automatic acoustic bat identification system based on the audible spectrum | |
CN115578678A (en) | Fish feeding intensity classification method and system | |
Wang et al. | A lightweight CNN-based model for early warning in sow oestrus sound monitoring | |
NL2029582B1 (en) | Method for intelligently identifying estrus state through sheep sound | |
Poorjam et al. | Quality control of voice recordings in remote Parkinson’s disease monitoring using the infinite hidden Markov model | |
Waghmare et al. | Development of isolated marathi words emotional speech database | |
Zhang et al. | Environmental sound recognition using double-level energy detection | |
Wang et al. | Identification of oestrus cows based on vocalisation characteristics and machine learning technique using a dual-channel-equipped acoustic tag | |
Du et al. | A tristimulus-formant model for automatic recognition of call types of laying hens | |
CN115762578A (en) | Interpretable heart sound abnormity identification method and system based on fractional domain Fourier transform | |
Ntalampiras et al. | Automatic detection of cow/calf vocalizations in free-stall barn | |
Zhang et al. | A novel insect sound recognition algorithm based on mfcc and cnn | |
CN114155833A (en) | Neural network-based baby cry identification method | |
CN113823295A (en) | Method for intelligently identifying oestrus state through sheep voice | |
Özmen et al. | Sound analysis to recognize cattle vocalization in a semi-open barn | |
Zhou et al. | Environmental sound classification of western black-crowned gibbon habitat based on subspace method and DenseNet121 | |
CN115206347A (en) | Method and device for identifying bowel sounds, storage medium and computer equipment | |
Shi | An Improved Machine Learning Model for Pig Abnormal Voice Recognition | |
CN117727330B (en) | Biological diversity prediction method based on audio decomposition |