CN115413605B - System for discriminating sex of meat pigeons by combining weight, sound and struggling force information - Google Patents
System for discriminating sex of meat pigeons by combining weight, sound and struggling force information Download PDFInfo
- Publication number
- CN115413605B CN115413605B CN202211136861.3A CN202211136861A CN115413605B CN 115413605 B CN115413605 B CN 115413605B CN 202211136861 A CN202211136861 A CN 202211136861A CN 115413605 B CN115413605 B CN 115413605B
- Authority
- CN
- China
- Prior art keywords
- module
- sound
- pigeons
- data
- weight
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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
- A01K45/00—Other aviculture appliances, e.g. devices for determining whether a bird is about to lay
Abstract
The application provides a system for distinguishing gender of meat pigeons by integrating weight, sound and struggling force information, which comprises a base, a weight module, a bracket, a sound module, a struggling force module and a data terminal module; the weight module comprises a weighing sensor, the sound module comprises a microphone probe, the struggling force module comprises a fixed plate, a connecting block and two pairs of mechanical clamping claws, and a plurality of patch type pressure sensors are uniformly arranged on the inner side walls of the mechanical clamping claws; the weight module, the sound module and the struggling force module are respectively and remotely connected with the data terminal module. The system integrates the weight, sound and struggling force information of the pigeons to carry out automatic sex judgment of the pigeons, does not need to be identified by human intervention, has small human influence factors and small errors, and has high judgment accuracy; meanwhile, the system is simple to operate, short in time consumption and high in identification and authentication efficiency.
Description
Technical Field
The invention relates to the technical field of poultry farming, in particular to a system for judging the sex of meat pigeons by integrating weight, sound and struggling force information.
Background
The meat pigeon breeding technology is a technical means for scientific breeding, disease protection and the like of meat pigeons by using various disciplinary measures through an artificial breeding method; the meat pigeon contains more than 17 amino acids, the total amino acids reaches 53.9%, and contains more than 10 microelements and multiple vitamins, so that the meat pigeon has rich nutritive value.
In the meat pigeon breeding process, male sex and female sex of the meat pigeons are usually required to be judged, so that the meat pigeons with different sexes are placed in the same cage to carry out spouse and spawning, and the yield and scale of meat pigeon breeding are increased. However, unlike traditional poultry such as chickens and ducks, the different meat pigeons have small differences in appearance characteristics, and the same-age meat pigeons among different types are basically consistent in appearance, so that the sex of the meat pigeons cannot be rapidly, effectively and accurately distinguished directly from the appearance of the meat pigeons. In the prior art, the sex judgment of the meat pigeons is generally carried out by the breeder through the factors of the appearance, sound, behavior, hand-caught struggling force and the like of the meat pigeons, however, the artificial judgment needs to have rich identification experience, and the artificial subjective influence factors are high, the error is large and the accuracy is low. Meanwhile, the sex judgment of the pigeons is carried out by the methods of anus identification, pelvis measurement, gene sequencing and the like in the prior art, but the anus identification has great stimulation to the pigeons and is easy to influence the subsequent spouse and quality of the pigeons, the pelvis vehicle is only suitable for female after egg laying, the application range is narrow, the gene sequencing identification cost is high, the time is long, and the pigeons produced in a large scale and in industrialization cannot be used.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention aims to provide a system for judging the sex of the pigeons by integrating weight, sound and struggling force information, which is used for automatically judging the sex of the pigeons by integrating the weight, sound and struggling force information of the pigeons, does not need to be identified by human, has small human influence factors and errors and has high judgment accuracy; meanwhile, the system is simple to operate, short in time consumption and high in identification and authentication efficiency.
The aim of the invention is achieved by the following technical scheme:
a system for discriminating sex of meat pigeons by combining weight, sound and struggling force information is characterized in that: the device comprises a base, a weight module, a bracket, a sound module, a struggling force module and a data terminal module; the weight module is fixedly arranged on the end face of the base, a bracket is arranged on the weight module, and the weight module comprises a weighing sensor; the middle part of the bracket is provided with a sound module which comprises a microphone probe, and the upper end part of the bracket is provided with a struggling force module; the struggling force module comprises a fixed plate, a connecting block and two pairs of mechanical clamping claws, wherein the fixed plate is arranged at the upper end of the bracket, the end face of the fixed plate, which is far away from the bracket, is provided with the connecting block, the end of the connecting block, which is far away from the fixed plate, is provided with the two pairs of mechanical clamping claws, the two pairs of mechanical clamping claws are mutually parallel, and the claw inner side walls of the two pairs of mechanical clamping claws are respectively and uniformly provided with a plurality of patch type pressure sensors; the weight module, the sound module and the struggling force module are respectively and remotely connected with the data terminal module.
A method for discriminating sex of meat pigeons by combining weight, sound and struggling force information is characterized in that the method is adopted by the system and is characterized in that: the method specifically comprises the following steps:
step S001: initializing parameters of a weight module, a sound module, a struggling force module and a data terminal module;
step S002: two pairs of mechanical clamping claws are adopted to clamp the pigeons to be identified, and the clamping method comprises the following steps: the position for grabbing the pigeons is a pair of wings, namely the wings are placed in the area of the patch type pressure sensor and firmly clamped;
step S003: setting an acquisition time interval of a weighing sensor to obtain the actual weight of the pigeon;
step S004: loosening the pigeons, initializing the weight module, then re-clamping the pigeons by adopting the method of the step S002, simultaneously starting the weight module, the sound module and the struggling force module, acquiring a Mel spectrogram of the data fluctuation rate, the pressure sensing data fluctuation rate and the sound information of the weighing sensor according to the feedback information of the weight module, the sound module and the struggling force module, and carrying out preliminary judgment on the probability that the sound information is male or female through the Mel spectrogram;
step S005: the data terminal module takes the data fluctuation rate of the weighing sensor, the data fluctuation rate of the pressure sensor and the probability of male or female through a Mel spectrum in the step S004 as the input of a support vector machine, and the output of the sex identification of the meat pigeons is obtained.
Further preferably, the step S003 is specifically: first set upAcquisition interval time delta t of weighing sensor 1 I.e. every Δt 1 Collecting primary data in a time period; when the mechanical clamping claw clamps the pigeons, the current time is collected and recorded to be shifted backwards by n 1 Weighing sensor data of the sampling points to obtain n 1 Maximum value m of data of each sampling point max Minimum value m min And average value m p And calculating the ratio N of the difference between the maximum value and the minimum value to the average value, i.e
When N is less than 0.5%, judging that the pigeon is in a static state without struggling, wherein N is 1 Average value m of data of each sampling point p The actual weight of the meat pigeon; if N is greater than or equal to 0.5%, initializing a weight module and re-measuring until the actual weight of the meat pigeon is obtained; the actual weight of the pigeons can also be measured according to the method described above, while the held pigeons are in an un-struggled state (visual inspection).
The specific method for obtaining the data fluctuation rate of the weighing sensor in the step S004 is as follows:
step S411: resetting the acquisition interval delta t of the weighing sensor 1 Simultaneously, the acquisition time t of the weighing sensor in the struggling state of the meat pigeons is set 1 ；
Step S412: calculating to obtain n 2 The data fluctuation rate P of the weighing sensor at each sampling point is specifically as follows:
wherein: m is m 0 Is the actual weight of the meat pigeon; m is m i The actual test value of the ith sampling point of the weighing sensor;
step S413: presetting a weighing sensor data fluctuation rate threshold value P 0 By a backward shift of the current time by n 2 Load cell data for individual sampling pointsFluctuation rate P and threshold P 0 Comparing to judge whether the pigeons are in struggling state;
if P > P 0 Judging that the pigeons placed on the weighing sensor are in a struggling state, and recording the whole acquisition time t 1 The data fluctuation rate of the inner weighing sensor; if P is less than or equal to P 0 Judging that the pigeons placed on the weighing sensor are not in a struggling state;
step S414: and storing the acquired data, and initializing the measurement data of the weight module so as to facilitate the next measurement.
The specific method for obtaining the data fluctuation rate of the pressure sensor in the step S004 is as follows:
step S421: acquisition interval time delta t for setting patch type pressure sensor 2 I.e. every Δt 2 Collecting once in a time period;
step S422: receiving the struggle signal of meat pigeons, collecting and recording the current time of each patch type pressure sensor and then shifting n 3 Sampling data of the sampling points, and n is as follows 3 The average value of the sampling points is taken as the pressure value Q0 of the non-struggling patch type pressure sensor; removing the non-contact patch type pressure sensor according to the zero pressure of any collecting time period, recording the data of the contact patch type pressure sensor, and recording the number; acquisition time of patch type pressure sensor and acquisition time t of weighing sensor in weight module 1 The same;
step S423: the data fluctuation rate Q of each patch type pressure sensor is obtained through calculation, and specifically comprises the following steps:
wherein: q (Q) i Is the ith pressure value of the patch type pressure sensor.
The specific method for obtaining the mel spectrogram of the sound information in step S004 is as follows:
step S431: presetting a frame time in a sound module to be longThe length of the time and the frame shift time are set at the same time, and the time length t of the sound information sample is set 2 ；
Step S432: judging whether the pigeons are positioned on the device or not according to the data fed back by the weight module; if yes, starting a sound module to collect sound information;
step S433: the collected sound information is transmitted to a data terminal module through remote transmission, the data terminal module carries out pre-processing of aggravating and denoising on the sound information firstly, then a single-parameter double-threshold sound endpoint detection method with short-time energy is used for identifying a sound information section of the meat pigeon, silent parts in the collected sound information are removed, and finally whether the sound information is effective or not, namely whether the sound of the meat pigeon exists or not is judged; if not, the sound information is collected and processed again;
step S434: the effective sound information processed in step S433 is processed according to the set sample time length t 2 And converting into a mel-pattern by using a mel filter.
Further optimizing, the preprocessing of the sound information in the step S433 specifically includes the steps of:
firstly, emphasizing a high-frequency part of sound information and increasing the high-frequency resolution of the sound information of the meat pigeons, wherein the specific formula is as follows:
y (n) ＝x (n) -ax (n-1) ；
wherein: y is (n) The sound signal acquisition point of the meat pigeon after the n pre-emphasis; x is x (n) The sound signal acquisition point of the nth meat pigeon; x is x (n-1) The sound signal acquisition point of the (n-1) th meat pigeon; a is a fixed value, and is generally 0.9-1;
and then removing the environmental additive noise by adopting a spectrum subtraction noise method, so that the interference of the mutual superposition of the environmental noise on the sound information of the meat pigeons is reduced.
Further optimizing, the step S433 of identifying the voice information section of the pigeon by using the single-parameter double-threshold voice endpoint detection method with short-time energy comprises the following specific steps:
first, short-term energy E i And short-time zero-crossing rate Z i Determining two thresholds;
wherein the short-term energy E i The method comprises the following steps:
wherein: m is the frame length, namely the sampling point length of each frame after the sound is divided into frames; n represents the sampling point serial number corresponding to each frame of the pigeon sound, and i is the frame serial number;
x i (n) is the normalized amplitude of the sound signal, which is specifically:
wherein: x' (n) is the original sound signal, n is the sampling point number;
short time zero crossing rate Z i The method comprises the following steps:
wherein: m is the frame length, namely the sampling point length of each frame after the sound is divided into frames; n represents the sampling point serial number corresponding to each frame of the pigeon sound, and i is the frame serial number; x is x i (n) normalizing the amplitude for the sound signal;
when at the sample time length t 2 In the presence of interval E i ＞E m And N is i ＞N m When the pigeon voice information interval is detected; when the length of the meat pigeon sound information interval is less than the sample time length t 2 When the signal is intercepted, one end of the signal is not sound, and the completion is the sample time length t 2 The method comprises the steps of carrying out a first treatment on the surface of the When the length of the meat pigeon sound information interval is longer than the sample time length t 2 When intercepting the last E in the interval i ＞E m And N is i ＞N m The time length of the sampling point at the time, therebyMaintaining the sound length and sample time length t of meat pigeons 2 And consistent.
The specific method for performing preliminary judgment on the probability of the sound information being male or female through the mel spectrogram in the step S004 is as follows:
firstly, sequentially carrying out three identical convolution modules on the transformed Mel spectrogram matrix for processing; each convolution module process comprises a convolution network of 3*3 twice, a rule activation function is used and standardized (BN) is carried out for each convolution, and the maximum pooling process with the size of 2 x 2 and the step length of 2 is carried out;
then, the output result is processed by a convolution network of four times 3*3, each convolution uses a rule activation function and performs standardization (BN) processing, and then the maximum pooling with the size of 2 x 2 and the step length of 2 is used for processing;
then, the output size is output after being adjusted, is used as the input of the 128-output-dimension full link layer FC1, and is subjected to normalization (BN) processing by using a rule activation function; then taking the output result as the input of the 3-output-dimension full-link layer FC2, and using a rule activation function to perform standardization (BN) processing; wherein, the 3 output dimension represents that the samples are of three types (namely adult female pigeon sound, adult male pigeon sound and non-meat pigeon sound);
and finally, inputting the final output result into a Softmax classifier to obtain a final classification result.
Further preferably, the step S005 specifically includes:
step S511: according to the data fluctuation rate P of the weighing sensor, the data fluctuation rate Q of the pressure sensor and the probabilities L and H of male or female of sound information through a Mel spectrogram, wherein L is the probability of primarily judging that the sound information is male through the Mel spectrogram, and H is the probability of primarily judging that the sound information is female through the Mel spectrogram, an input matrix X with four dimensions is obtained, namely:
X＝[P,Q,L,H]；
step S512: using a support vector machine as a classification model, and outputting a classification result of-1 or 1; wherein-1 represents that the pigeons are female, and 1 represents that the pigeons are male;
the data training set of the support vector machine is t= [ (X) 1 ,k 1 )，…,(X i ,k i ),…,(X n ,k n )]The method comprises the steps of carrying out a first treatment on the surface of the Wherein i is the number of training samples of the support vector machine, and X i Data for the ith training sample, k i The judgment result is the i training sample;
since the training samples are random and the obtained data are nonlinear, the decision function of the support vector machine is:
||X i ,X j || 2 ＝P i P j +Q i Q j +L i L j +H i H j ；
wherein: sigma is a constant;
and the decision function f (X) is judged according to the following principle:
finally, the judgment of the sex of the meat pigeon is completed according to the output result obtained by the input sample X of the decision function f (X).
The invention has the following technical effects:
the system has simple structure and convenient use, and combines the methods of mechanical learning, convolutional neural network, support vector machine and the like to identify the sex of the pigeons, so that the information of the weight, struggling force, sound and the like of the pigeons is effectively integrated to judge the sex of the pigeons, thereby ensuring high accuracy and small error of judgment; meanwhile, the sex judgment of the pigeons is realized through an automatic program, the problems of large human influence factor, high labor intensity of operators, long time consumption of human judgment and the like in the human judgment process are avoided, and therefore the quick and accurate judgment is realized, the judgment efficiency is improved, and the pairing rate in the pigeon breeding process is increased. In addition, the method reduces the identification time of the sex of the pigeons, can greatly shorten the pairing time of the pigeons and improves the spawning rate and yield of the pigeons.
Drawings
Fig. 1 is a schematic structural diagram of a system for discriminating sex of meat pigeons in an embodiment of the present invention.
FIG. 2 is a flow chart of a method for determining sex of meat pigeons in an embodiment of the invention.
Fig. 3 is a schematic diagram of a meat pigeon after acoustic denoising in an embodiment of the present invention.
FIG. 4 is a graph of short-term energy of sound of a pigeon in an embodiment of the invention.
Fig. 5 is a graph of a short time zero crossing rate of sound of a pigeon in an embodiment of the invention.
10, a base; 20. a weight module; 30. a bracket; 40. a sound module; 50. a struggling force module; 51. a fixing plate; 52. a connecting block; 53. mechanical gripper jaws; 530. a patch type pressure sensor.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments.
Example 1:
as shown in fig. 1, a system for discriminating sex of meat pigeons by integrating weight, sound and struggling force information is characterized in that: comprises a base 10, a weight module 20, a bracket 30, a sound module 40, a struggling force module 50 and a data terminal module; the weight module 20 is fixedly arranged on the end face of the base 10, a bracket 30 is arranged on the weight module 20, and the weight module 20 comprises a weighing sensor; the middle part of the bracket 30 is provided with a sound module 40, the sound module 40 comprises a microphone probe, and the upper end part of the bracket 30 is provided with a struggling force module 50; the struggling force module 50 comprises a fixed plate 51, a connecting block 52 and two pairs of mechanical clamping claws 53, wherein the fixed plate 51 is arranged at the upper end of the bracket 30, the connecting block 52 is arranged at the end face of one end of the fixed plate 51 far away from the bracket 30, two pairs of mechanical clamping claws 53 are arranged at one end of the connecting block 52 far away from the fixed plate 31 and are mutually parallel (as shown in fig. 1, a single mechanical clamping claw 53 is rotationally connected with the connecting block 52), a plurality of patch type pressure sensors 530 are uniformly arranged on the inner side walls of the two pairs of mechanical clamping claws 53 respectively (as shown in fig. 1, four patch type pressure sensors 530 are arranged in the single mechanical clamping claw 53, namely 16 patch type pressure sensors 530 are arranged on the inner walls of the two groups of mechanical clamping claws 53 in total); the weight module 20, the sound module 40 and the struggling force module 50 are respectively and remotely connected with the data terminal module.
The mechanical clamping claw 53 can accurately regulate and control the clamping force by an electric or starting device, so that damage to the pigeon during the clamping process is avoided.
Example 2:
as shown in fig. 2 to 5, a method for discriminating sex of meat pigeons by integrating weight, sound and struggle force information, using the system as described in the above embodiment 1, is characterized in that: the method specifically comprises the following steps:
step S001: initializing parameters of a weight module, a sound module, a struggling force module and a data terminal module;
step S002: two pairs of mechanical clamping claws are adopted to clamp the pigeons to be identified, and the clamping method comprises the following steps: the position for grabbing the pigeons is a pair of wings, namely the wings are placed in the area of the patch type pressure sensor and firmly clamped;
step S003: setting an acquisition time interval of a weighing sensor to obtain the actual weight of the pigeon; the method comprises the following steps:
firstly, setting acquisition interval time delta t of a weighing sensor 1 (in the present embodiment, Δt) 1 =5 ms), i.e. data is collected once every 5ms period; when the mechanical gripper jaw 53 grips the meat pigeon, the current time lapse n is collected and recorded 1 (n 1 Load cell data of =20) sampling points, 20 samples were obtainedMaximum value m of point data max Minimum value m min And average value m p And calculating the ratio N of the difference between the maximum value and the minimum value to the average value, i.e
When N is less than 0.5%, judging that the pigeon is in a static state without struggling, and at the moment, averaging the data of 20 sampling points to obtain an average value m p The actual weight of the meat pigeon; if N is greater than or equal to 0.5%, initializing a weight module and re-measuring until the actual weight of the meat pigeon is obtained; the actual weight of the pigeons can also be measured according to the method described above, while the held pigeons are in an un-struggled state (visual inspection).
Step S004: loosening the pigeons, initializing the weight module, then re-clamping the pigeons by adopting the method of the step S002, simultaneously starting the weight module, the sound module and the struggling force module, acquiring a Mel spectrogram of the data fluctuation rate of the weighing sensor, the data fluctuation rate of the pressure sensor and the sound information according to the feedback information of the weight module, the sound module and the struggling force module, and carrying out preliminary judgment on the probability that the sound information is male or female through the Mel spectrogram;
the specific method for acquiring the data fluctuation rate of the weighing sensor comprises the following steps:
step S411: resetting the acquisition interval delta t of the weighing sensor 1 (in the present embodiment, Δt) 1 =5 ms), and simultaneously sets the acquisition time t of the weighing sensor in the struggling state of the meat pigeons 1 (in this embodiment, t) 1 ＝1s)；
Step S412: calculating to obtain n 2 (n 2 =20) load cell data fluctuation rate P at sampling points, specifically:
wherein: m is m 0 Is the actual body of meat pigeonWeighing; m is m i The actual test value of the ith sampling point of the weighing sensor;
step S413: presetting a weighing sensor data fluctuation rate threshold value P 0 Load cell data fluctuation rate P and threshold P by backward shifting current time by 20 sampling points 0 A comparison is made (the threshold of the ripple rate is obtained from a conventional large number of test data, in this example,
) Thereby judging whether the meat pigeons are in struggling state;
if P > P 0 Judging that the meat pigeons placed on the weighing sensor are in a struggling state, and recording the data fluctuation rate of the weighing sensor within 1s of the whole acquisition time; if P is less than or equal to P 0 Judging that the pigeons placed on the weighing sensor are not in a struggling state;
step S414: and storing the acquired data, and initializing the measurement data of the weight module so as to facilitate the next measurement.
The specific method for acquiring the data fluctuation rate of the pressure sensor comprises the following steps:
step S421: acquisition interval time delta t for setting patch type pressure sensor 2 (in this embodiment, Δt) 2 =5 ms), i.e. once every 5ms period;
step S422: receiving the struggle signals of meat pigeons, simultaneously collecting data of 16 patch type pressure sensors and recording the current time of each patch type pressure sensor for n times 3 (n 3 =20) sampling data of sampling points, taking the average value of 20 sampling points as the pressure value Q of the non-struggle patch type pressure sensor 0 The method comprises the steps of carrying out a first treatment on the surface of the Removing a non-contact patch type pressure sensor according to the fact that the pressure is zero in any collecting time period, recording data of the contact patch type pressure sensor, and recording a number; acquisition time of patch type pressure sensor and acquisition time t of weighing sensor in weight module 1 The same, namely 1s;
step S423: the data fluctuation rate Q of each patch type pressure sensor is obtained through calculation, and specifically comprises the following steps:
wherein: q (Q) i Is the ith pressure value of the patch type pressure sensor.
The specific method for acquiring the mel spectrogram of the sound information comprises the following steps:
step S431: presetting a frame dividing time length and a frame shifting time length in a sound module, and simultaneously setting a sound information sample time length t 2 (t 2 ＞t 1 And in this embodiment: t is t 2 ＝3s)；
The frame time length and the frame shift time length are set to satisfy the size of the mel spectrogram set later, for example: the sampling frequency of the sound signal is 32000Hz, the sampling time is 3s, the sampling data points are 3 x 32000, the frame time length is 35ms, and the frame shift time length is 9.25ms
(the number of sound frames finally obtained needs to be rounded) and the number of points of each frame can be 35×32=1120, namely 2048-point real-sequence fast fourier transform (rfft) can be selected, and the number of the fast fourier transform points is generally required to be an integer power of 2. Since the sound signal data is basically symmetrical, only a non-negative half frequency, namely 16000Hz, is selected, a matrix (320, 1024+1) is obtained after real-sequence fast fourier transformation (1 is added to the matrix to be a point corresponding to 16000 Hz), and the (320, 1024+1) matrix is multiplied by a transpose of the 128-dimensional mel filter data matrix (128, 1024+1) to obtain a mel map data matrix (320, 128), namely the size of the mel map obtained by final transformation is 320×128.
Step S432: judging whether the pigeons are positioned on the device or not according to the data fed back by the weight module; if yes, starting a sound module to collect sound information;
step S433: the collected sound information is transmitted to the data terminal module through the remote transmission,
the data terminal module firstly carries out the preprocessing of aggravating and denoising on the sound information, and specifically comprises the following steps:
firstly, emphasizing a high-frequency part of sound information and increasing the high-frequency resolution of the sound information of the meat pigeons, wherein the specific formula is as follows:
y (n) ＝x (n) -ax (n-1) ；
wherein: y is (n) The sound signal acquisition point of the meat pigeon after the n pre-emphasis; x is x (n) The sound signal acquisition point of the nth meat pigeon; x is x (n-1) The sound signal acquisition point of the (n-1) th meat pigeon; a is a constant value, and is generally 0.9 to 1 (preferably 0.95 in this embodiment);
then removing environmental additive noise by adopting a spectrum subtraction noise method, so as to reduce interference of mutual superposition of the environmental noise on sound information of the meat pigeons;
such as: sound signal y (n) From the sound signal x of pigeons (n) And additive noise d (n) Composition, i.e. y (n) ＝x (n) +d n( After fourier transformation, the method comprises the following steps: y is Y (w) ＝X (w) +D (w) (wherein Y (w) Is y (n) Fourier transform, X of (X) (w) Is x (n) Fourier transform, D (w) Is d (n) Fourier of (a) transform),
at the same time:
wherein: k represents the sampling frame number corresponding to the sound signal of the meat pigeon, Y i (ω) represents a silence signal corresponding to k sampling frames;
the original sound signal spectrum subtracting formula is:
wherein: a represents an over-subtraction factor, defaults to 5, b represents a gain compensation factor, defaults to 0.002;
x is to be i (omega) performing Fourier transform to obtain an i-th frame pigeon sound signal x i(n) Thereby reducing the sound signal of the meat pigeon caused by the mutual superposition of the environmental noiseDisturbance of the message.
Then, a single-parameter double-threshold sound endpoint detection method with short-time energy is used for identifying a sound information interval of the meat pigeon, and the method specifically comprises the following steps:
first, short-term energy E i And short-time zero-crossing rate Z i Determining two thresholds;
wherein the short-term energy E i The method comprises the following steps:
wherein: m is the frame length, namely the sampling point length of each frame after the sound is divided into frames; n represents the sampling point serial number corresponding to each frame of the pigeon sound, and i is the frame serial number;
x i (n) is the normalized amplitude of the sound signal, which is specifically:
wherein: x' (n) is the original sound signal, n is the sampling point number;
short time zero crossing rate Z i The method comprises the following steps:
wherein: m is the frame length, namely the sampling point length of each frame after the sound is divided into frames; n represents the sampling point serial number corresponding to each frame of the pigeon sound, and i is the frame serial number; x is x i (n) normalizing the amplitude for the sound signal;
when at the sample time length t 2 In the presence of interval E i ＞E m And N is i ＞N m When the pigeon voice information interval is detected; when the length of the meat pigeon sound information interval is less than the sample time length3s, intercepting one end of the no-sound signal and completing the no-sound signal to be the sample time length t 2 The method comprises the steps of carrying out a first treatment on the surface of the When the length of the meat pigeon sound information interval is greater than the sample time length for 3s, the last E in the interception interval i ＞E m And N is i ＞N m The time length of the sampling point is equal to the time length of the sample, so that the sound length of the meat pigeon is consistent with the time length of the sample for 3 s.
Removing silent parts in the collected sound information, and finally judging whether the sound information is effective, namely whether the sound of the pigeon exists; if not, the sound information is collected and processed again; the method comprises the following steps:
obtaining a pigeon sound information section according to the operation, wherein the pigeon sound in the section has a voiced sound part and an unvoiced sound part (removing silence and noise), the voiced sound part and the unvoiced sound part are sound information required to be collected, and the unvoiced sound part belongs to consonants in the sound and has small energy, and the energy of the voiced sound is higher than that of the unvoiced sound; furthermore, the main sound information required in the present application is the voiced sound part of pigeon sound, so a higher short-time energy is taken as a threshold, e.g. E lim =15, and the duration of each pronunciation is typically greater than 100ms according to the sound characteristics of the meat pigeon pronunciation. When the short-time energy curve exists E i ＞E lim And i is greater than 100/frame shift (i.e., frame number)>100/frame shift), it is determined that a voiced sound part exists in the pigeon sound information section, that is, that the pigeon sound exists. As shown in fig. 3, 4 and 5, the pigeon sound information interval can be obtained between about 0.6 and 2.3s according to the above operation, and then the voiced sound part is judged to exist between 0.6 and 1.8s through the energy curve, so that the interval is determined to be the effective sound information interval.
Step S434: the effective sound information processed in step S433 is processed according to the set sample time length t 2 I.e. 3s, is converted to a mel profile using a mel filter.
Preliminary judgment of male or female probability of sound information is carried out through a Mel spectrogram, specifically:
placing the mel spectrogram into a convolutional neural network model, which is known by the steps: the obtained mel-graph data matrix is (320,128), namely the network input size is 320×128;
firstly, sequentially carrying out three identical convolution modules on the transformed Mel spectrogram matrix for processing; each convolution module process comprises a convolution network of 3*3 twice, a rule activation function is used and standardized (BN) is carried out for each convolution, and the maximum pooling process with the size of 2 x 2 and the step length of 2 is carried out;
firstly, processing input data (320,128,1) through a first convolution module, namely, through a convolution network with 3*3 twice, wherein rule activation functions are used for each convolution and Batch Normalization (BN) is carried out, the output is 32, the output is 320,128,32, the maximum pooling with the size of 2 x 2 and the step length of 2 is used, and the final network output is 160,64,32; processing the data (160,64,32) through a second convolution module, namely, through a convolution network with 3*3 twice, wherein each convolution uses a rule activation function and performs Batch Normalization (BN), the output is 64, the output is 160,64,64, the maximum pooling with the size of 2 x 2 and the step length of 2 is used, and the final network output is 80,32,64; the data (80,32,64) is then processed through a third convolution module, namely through a convolution network of 3*3 twice, each convolution uses a rule activation function and performs Batch Normalization (BN), the output is characterized by 128, the output is (80,32,128), the maximum pooling with a size of 2 x 2 and a step size of 2 is used, and the final network output is (40,16,128).
Then, the output result (40,16,128) is processed by a four-time 3*3 convolution network, rule activation function is used for each convolution, normalization (BN) processing is carried out, the output is 128, the output is 80,32,128, the processing is carried out by using 2 x 2 maximum pooling with the step length of 2, and the final network output is 20,8,256;
then, the output size is adjusted to be output, namely, the output size (20,8,256) is adjusted to be (20,2048)
Taking the adjusted data (20,2048) as an input of the 128-output-dimension full link layer FC1, and performing normalization (BN) processing by using a rule activation function, wherein the output is (20, 128); then the output result is used as the input of the 3-output dimension full link layer FC2, a rule activation function is used for normalization (BN) processing, and the output is (20, 3); wherein, the 3 output dimension represents that the samples are of three types (namely adult female pigeon sound, adult male pigeon sound and non-meat pigeon sound);
and finally, inputting the final output result into a Softmax classifier to obtain a final classification result.
Step S005: the data terminal module takes the data fluctuation rate of the weighing sensor, the data fluctuation rate of the pressure sensor and the probability of male or female through a mel map in the step S004 as the input of a support vector machine, and the output of the sex identification of the meat pigeons is obtained specifically as follows:
step S511: according to the data fluctuation rate P of the weighing sensor, the data fluctuation rate Q of the pressure sensor and the probabilities L and H of male or female of sound information through a Mel spectrogram, wherein L is the probability of primarily judging that the sound information is male through the Mel spectrogram, and H is the probability of primarily judging that the sound information is female through the Mel spectrogram, an input matrix X with four dimensions is obtained, namely:
X＝[P,Q,L,H]；
step S512: using a support vector machine as a classification model, and outputting a classification result of-1 or 1; wherein-1 represents that the pigeons are female, and 1 represents that the pigeons are male;
the data training set of the support vector machine is t= [ (X) 1 ,k 1 )，…,(X i ,k i ),…,(X n ,k n )]The method comprises the steps of carrying out a first treatment on the surface of the Wherein i is the number of training samples of the support vector machine, and X i Data for the ith training sample, k i The judgment result is the i training sample;
since the training samples are random and the obtained data are nonlinear, the decision function of the support vector machine is:
||X i ,X j || 2 ＝P i P j +Q i Q j +L i L j +H i H j ；
wherein: sigma is a constant;
and the decision function f (X) is judged according to the following principle:
finally, the judgment of the sex of the meat pigeon is completed according to the output result obtained by the input sample X of the decision function f (X).
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned. Furthermore, it should be understood that although the present disclosure describes embodiments, not every embodiment is provided with a separate embodiment, and that this description is provided for clarity only, and that the disclosure is not limited to the embodiments described in detail below, and that the embodiments described in the examples may be combined as appropriate to form other embodiments that will be apparent to those skilled in the art.
Claims (3)
1. A method for distinguishing sex of meat pigeons by combining weight, sound and struggling force information is characterized in that: the system comprises a base, a weight module, a bracket, a sound module, a struggling force module and a data terminal module; the weight module is fixedly arranged on the end face of the base, a bracket is arranged on the weight module, and the weight module comprises a weighing sensor; the middle part of the bracket is provided with a sound module which comprises a microphone probe, and the upper end part of the bracket is provided with a struggling force module; the struggling force module comprises a fixed plate, a connecting block and two pairs of mechanical clamping claws, wherein the fixed plate is arranged at the upper end of the bracket, the end face of the fixed plate, which is far away from the bracket, is provided with the connecting block, the end of the connecting block, which is far away from the fixed plate, is provided with the two pairs of mechanical clamping claws, the two pairs of mechanical clamping claws are mutually parallel, and the claw inner side walls of the two pairs of mechanical clamping claws are respectively and uniformly provided with a plurality of patch type pressure sensors; the weight module, the sound module and the struggling force module are respectively and remotely connected with the data terminal module;
the method for judging the sex of the meat pigeons specifically comprises the following steps:
step S001: initializing parameters of a weight module, a sound module, a struggling force module and a data terminal module;
step S002: two pairs of mechanical clamping claws are adopted to clamp the pigeons to be identified, and the clamping method comprises the following steps: the position for grabbing the pigeons is a pair of wings, namely the wings are placed in the area of the patch type pressure sensor and firmly clamped;
step S003: setting an acquisition time interval of a weighing sensor to obtain the actual weight of the pigeon;
step S004: loosening the pigeons, initializing the weight module, then re-clamping the pigeons by adopting the method of the step S002, simultaneously starting the weight module, the sound module and the struggling force module, acquiring a Mel spectrogram of the data fluctuation rate of the weighing sensor, the data fluctuation rate of the pressure sensor and the sound information according to the feedback information of the weight module, the sound module and the struggling force module, and carrying out preliminary judgment on the probability that the sound information is male or female through the Mel spectrogram;
step S005: the data terminal module takes the data fluctuation rate of the weighing sensor, the data fluctuation rate of the pressure sensor and the probability of male or female through a Mel spectrum in the step S004 as the input of a support vector machine, and obtains the output of the sex identification of the meat pigeons;
the step S003 specifically includes: firstly, weighing is setAcquisition interval Δt of sensor 1 I.e. every Δt 1 Collecting primary data in a time period; when the mechanical clamping claw clamps the pigeons, the current time is collected and recorded to be shifted backwards by n 1 Weighing sensor data of the sampling points to obtain n 1 Maximum value m of data of each sampling point max Minimum value m min And average value m p And calculating the ratio N of the difference between the maximum value and the minimum value to the average value, i.e
When N is less than 0.5%, judging that the pigeon is in a static state without struggling, wherein N is 1 Average value m of data of each sampling point p The actual weight of the meat pigeon; if N is greater than or equal to 0.5%, initializing a weight module and re-measuring until the actual weight of the meat pigeon is obtained;
the specific method for acquiring the data fluctuation rate of the weighing sensor in the step S004 is as follows:
step S411: resetting the acquisition interval delta t of the weighing sensor 1 Simultaneously, the acquisition time t of the weighing sensor in the struggling state of the meat pigeons is set 1 ；
Step S412: calculating to obtain n 2 The data fluctuation rate P of the weighing sensor at each sampling point is specifically as follows:
wherein: m is m 0 Is the actual weight of the meat pigeon; m is m i The actual test value of the ith sampling point of the weighing sensor;
step S413: presetting a weighing sensor data fluctuation rate threshold value P 0 By a backward shift of the current time by n 2 Weighing sensor data fluctuation rate P and threshold P of each sampling point 0 Comparing to judge whether the pigeons are in struggling state;
if P > P 0 Judging that the pigeons placed on the weighing sensor are in a struggling state, and recording the whole acquisition time t 1 The data fluctuation rate of the inner weighing sensor; if P is less than or equal to P 0 Judging that the pigeons placed on the weighing sensor are not in a struggling state;
step S414: the acquired data are stored, and then the measurement data of the weight module are initialized so as to be convenient for the next measurement;
the specific method for acquiring the data fluctuation rate of the pressure sensor in the step S004 is as follows:
step S421: acquisition interval time delta t for setting patch type pressure sensor 2 I.e. every Δt 2 Collecting once in a time period;
step S422: receiving the struggle signal of meat pigeons, collecting and recording the current time of each patch type pressure sensor and then shifting n 3 Sampling data of the sampling points, and n is as follows 3 Average value of sampling points as pressure value Q of non-struggle patch type pressure sensor 0 The method comprises the steps of carrying out a first treatment on the surface of the Removing the non-contact patch type pressure sensor according to the zero pressure of any collecting time period, recording the data of the contact patch type pressure sensor, and recording the number; acquisition time of patch type pressure sensor and acquisition time t of weighing sensor in weight module 1 The same;
step S423: the data fluctuation rate Q of each patch type pressure sensor is obtained through calculation, and specifically comprises the following steps:
wherein: q (Q) i Is the ith pressure value of the patch type pressure sensor.
2. The method for determining sex of meat pigeons according to claim 1, wherein the method comprises the steps of: the specific method for acquiring the mel spectrogram of the sound information in step S004 includes:
step S431: in preset sound modulesThe frame dividing time length and the frame shifting time length are set at the same time, and the sound information sample time length t is set 2 ；
Step S432: judging whether the pigeons are positioned on the device or not according to the data fed back by the weight module; if yes, starting a sound module to collect sound information;
step S433: the collected sound information is transmitted to a data terminal module through remote transmission, the data terminal module carries out pre-processing of aggravating and denoising on the sound information firstly, then a single-parameter double-threshold sound endpoint detection method with short-time energy is used for identifying a sound information section of the meat pigeon, silent parts in the collected sound information are removed, and finally whether the sound information is effective or not, namely whether the sound of the meat pigeon exists or not is judged; if not, the sound information is collected and processed again;
step S434: the effective sound information processed in step S433 is processed according to the set sample time length t 2 And converting into a mel-pattern by using a mel filter.
3. The method for determining sex of meat pigeons according to claim 2, wherein the method comprises the steps of: the specific method for performing preliminary judgment on the probability that the sound information is male or female through the mel spectrogram in the step S004 is as follows:
firstly, sequentially carrying out three identical convolution modules on the transformed Mel spectrogram matrix for processing; each convolution module process comprises a convolution network of 3*3 twice, a rule activation function is used and standardized for each convolution, and the maximum pooling process with the size of 2 x 2 and the step length of 2 is carried out;
then, the output result is processed by a convolution network of four times 3*3, a rule activation function is used for each convolution, standardized processing is carried out, and the maximum pooling with the size of 2 x 2 and the step length of 2 is used for processing;
then, the output size is output after being adjusted, is used as the input of the 128-output-dimension full-link layer FC1, and is subjected to standardization processing by using a rule activation function; then taking the output result as the input of the 3-output-dimension full-link layer FC2, and using a rule activation function to perform standardization processing; wherein the 3 output dimensions represent samples of only three classes;
and finally, inputting the final output result into a Softmax classifier to obtain a final classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211136861.3A CN115413605B (en) | 2022-09-19 | 2022-09-19 | System for discriminating sex of meat pigeons by combining weight, sound and struggling force information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211136861.3A CN115413605B (en) | 2022-09-19 | 2022-09-19 | System for discriminating sex of meat pigeons by combining weight, sound and struggling force information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115413605A CN115413605A (en) | 2022-12-02 |
| CN115413605B true CN115413605B (en) | 2023-05-09 |
Family
ID=84203349
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211136861.3A Active CN115413605B (en) | 2022-09-19 | 2022-09-19 | System for discriminating sex of meat pigeons by combining weight, sound and struggling force information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115413605B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115420356B (en) * | 2022-09-13 | 2023-08-08 | 仲恺农业工程学院 | Sex identification method for adult pigeons |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109009129B (en) * | 2018-08-20 | 2019-06-04 | 南京农业大学 | Sow respiratory disease early warning system based on acoustic analysis |
| CN110367197B (en) * | 2019-08-26 | 2022-06-28 | 浙江省农业科学院 | Hybridization matching combination research method of meat pigeons |
| CN210295148U (en) * | 2019-09-20 | 2020-04-10 | 四川省畜牧科学研究院 | Image recognition workbench for identifying sex by turning anus of young poultry |
| JP7445956B2 (en) * | 2019-12-27 | 2024-03-08 | 有限会社電マーク | Information processing method, system and program for sexing chicken chicks, and method for sexing chicken chicks |
| JP2021106549A (en) * | 2019-12-27 | 2021-07-29 | 有限会社電マーク | Information processing method, system and program for sexing chick, and method for sexing chick |
| CN111461252A (en) * | 2020-04-13 | 2020-07-28 | 中国地质大学（武汉） | Chick sex detector and detection method |
| CN112997962A (en) * | 2021-03-18 | 2021-06-22 | 上海市农业科学院 | Assorted seed production method for Huayu white skin meat pigeon |
| CN113780408A (en) * | 2021-09-09 | 2021-12-10 | 安徽农业大学 | Live pig state identification method based on audio features |
| CN115048984A (en) * | 2022-05-17 | 2022-09-13 | 天津农学院 | Sow oestrus recognition method based on deep learning |
-
2022
- 2022-09-19 CN CN202211136861.3A patent/CN115413605B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115413605A (en) | 2022-12-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105841797B (en) | A kind of Car window electric machine extraordinary noise detection method and device based on MFCC and SVM | |
| CN109065027B (en) | Voice distinguishing model training method and device, computer equipment and storage medium | |
| CN115413605B (en) | System for discriminating sex of meat pigeons by combining weight, sound and struggling force information | |
| CN110044472B (en) | Intelligent detection system for abnormal sound and abnormal sound of online product | |
| CN108198574A (en) | Change of voice detection method and device | |
| CN109817227B (en) | Abnormal sound monitoring method and system for farm | |
| CN108694953A (en) | A kind of chirping of birds automatic identifying method based on Mel sub-band parameter features | |
| CN112990082B (en) | Detection and identification method of underwater sound pulse signal | |
| CN105916090B (en) | A kind of hearing aid device system based on intelligent speech recognition technology | |
| CN111508517A (en) | Intelligent micro-motor product control method based on noise characteristics | |
| CN112331231A (en) | Broiler feed intake detection system based on audio technology | |
| CN110897634A (en) | Electrocardiosignal generation method based on generation countermeasure network | |
| CN106340299A (en) | Speaker recognition system and method in complex environment | |
| CN110082106B (en) | Bearing fault diagnosis method based on Yu norm deep measurement learning | |
| CN115420356B (en) | Sex identification method for adult pigeons | |
| CN111352075B (en) | Underwater multi-sound-source positioning method and system based on deep learning | |
| JP4931927B2 (en) | Signal recognition method using a microcontroller | |
| JP6103591B2 (en) | Auscultation heart sound signal processing method, auscultation heart sound signal processing apparatus, and program for processing auscultation heart sound signal | |
| CN109919101A (en) | A kind of user experience evaluation method and system based on cell phone client | |
| CN108447504A (en) | A kind of oral evaluation method based on brain wave | |
| CN111639519A (en) | Living body detection method and device | |
| CN112735468A (en) | MFCC-based automobile seat motor abnormal noise detection method | |
| Liu et al. | Automatic terahertz recognition of hidden defects in layered polymer composites based on a deep residual network with transfer learning | |
| CN112185396B (en) | Offshore wind farm biological monitoring method and system based on passive acoustics | |
| Shi | An Improved Machine Learning Model for Pig Abnormal Voice Recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |