Method for embedding and extracting robust invisible text watermark

Abstract

The invention discloses a robust invisible text watermark embedding and extracting method, which comprises the following steps of: S1, segmenting a text carrier into which a watermark is to be embedded, and enabling each field to sequentially correspond to one bit of a bit string into which the watermark is embedded; s2, judging whether each bit of the bit string embedded with the watermark is 1 or 0, if the bit is 1, modifying a character at a specified position of a field corresponding to the bit into a character of which the initial is a letter before a preset letter, and keeping the semantics of the modified field unchanged; and if the bit is 0, modifying the character at the specified position of the field corresponding to the bit into the character of which the initial is not the letter before the preset letter, and keeping the semantics of the modified field unchanged. According to the method, the robustness of watermark embedding of the plain text can be improved.

Description

Technical Field

[0001] The present invention relates to the field of text watermarking, and in particular to a robust invisible text watermark embedding and extraction method. Background Art

[0002] Currently, text watermarks are mostly targeted at English texts, or long texts, or formatted texts such as PDF / Doc. There is no watermark solution specifically for pure Chinese characters and short Chinese sentences.

[0003] Some texts on the Internet are leaked due to confidentiality. For example, a doctor leaks text information about a patient's medical condition to a public network outside the internal medical network. Some special texts (such as confidential or internal texts) are leaked to the public network. In this case, it is necessary to track down the leaker. Summary of the Invention

[0004] The main purpose of the present invention is to provide a robust invisible text watermark embedding and extraction method, which is convenient for tracing the publisher of a short text in pure Chinese characters.

[0005] The technical solution adopted in the present invention is: A robust invisible text watermark embedding method is provided, comprising the following steps: S1. Segment the text carrier, and each field corresponds to a bit of the bit string embedded in the watermark; S2. Determine whether each bit of the bit string embedded with the watermark is 1 or 0. If it is 1, modify the word at the specified position of the field corresponding to the bit to a word whose first letter is the letter before the preset letter, and the semantics of the field remains unchanged after the modification; if it is 0, modify the word at the specified position of the field corresponding to the bit to a word whose first letter is not the letter before the preset letter, and the semantics of the field remains unchanged after the modification.

[0006] Following the above technical solution, alternatively, step S2 is specifically as follows: determining whether each bit of the bit string embedded with the watermark is 1 or 0; if it is 1, modifying the word at the specified position of the field corresponding to the bit to have an odd or even number of strokes, and the semantics of the field after modification remain unchanged; if it is 0, modifying the word at the specified position of the field corresponding to the bit to have an even or odd number of strokes, and the modification is exactly opposite to the modification when the bit is 1, and the semantics of the field after modification remain unchanged.

[0007] Following the above technical solution, alternatively, step S2 is specifically as follows: determining whether each bit of the bit string embedded with the watermark is 1 or 0; if it is 1, modifying the number of Chinese characters in the field corresponding to the bit to an odd number or an even number, and the semantics of the field remain unchanged after the modification; if it is 0, modifying the number of Chinese characters in the field corresponding to the bit to an even number or an odd number, and the modification is exactly opposite to the modification when the bit is 1, and the semantics of the field remain unchanged after the modification.

[0008] Following the above technical solution, when the number of Chinese characters is modified to an odd number or an even number, the modification method is to remove or add function words.

[0009] Following the above technical solution, the designated position is all the characters in the field, or the Chinese characters in the first half or the second half of the sub-segment, or the first character or the last character of the field.

[0010] Following the above technical solution, the text carrier is segmented specifically as follows: the text carrier is evenly divided into multiple segments according to the number of bit strings embedded with the watermark.

[0011] Following the above technical solution, the text carrier is segmented specifically as follows: segmenting according to the punctuation marks of the text carrier.

[0012] The present invention also provides a robust invisible text watermark extraction method, comprising the following steps: S1. Segment the text carrier, and each field corresponds to a bit of the bit string embedded in the watermark; S2. Determine in turn whether the first letter of the word at the specified position of each field is the letter before the preset letter. If so, the bit embedded in the field is 1; if not, the bit embedded in the field is 0.

[0013] Following the above technical solution, alternatively, step S2 is specifically as follows: determining in turn whether the number of Chinese characters in each field is odd or even; if it is odd, the bit embedded in the field is 1; if it is even, the bit embedded in the field is 0.

[0014] Following the above technical solution, alternatively, step S2 is specifically as follows: determining in turn whether the number of Chinese character strokes at the designated position of each field is odd or even; if it is odd, the bit embedded in the field is 1; if it is even, the bit embedded in the field is 0.

[0015] The beneficial effects of the present invention are as follows: the present invention can embed an invisible watermark in a short plain text according to the characteristics of Chinese characters. Specifically, the text carrier to be embedded with the watermark is segmented and the bit positions of the embedded bit string are modified in each field according to the bit positions of the embedded bit string, including modifying the character at a specified position in the field to a character whose first letter is the letter before a preset letter, or modifying the character at a specified position in the field to a character with an odd or even number of strokes, or modifying the number of Chinese characters in the field to an odd or even number, and the semantics of the modified field remain unchanged. Similarly, the watermark information can be extracted according to the above method. Therefore, the invisible watermark embedding method of the present invention can resist the copying, modification, deletion or rearrangement of the text. That is, even if the text is partially deleted or the sentence is rearranged, the watermark information can be extracted to a large extent.

[0016] Furthermore, when the number of words in the modified field is odd or even, the purpose of not changing the semantics of the field can be achieved by removing or adding function words.

[0017] Furthermore, the designated position in the field can be selected according to an agreement and is not restricted to a certain position, making the position of embedding the watermark more flexible and changeable.

[0018] Furthermore, the text segmentation can also be set as needed, without being restricted to a certain segmentation method, so that the field embedded in the watermark is variable.

[0019] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 This is the process of a robust invisible text watermark embedding method according to an embodiment of the present invention. Figure 1 ; Figure 2 This is the process of a robust invisible text watermark embedding method according to an embodiment of the present invention. Figure 2 ; Figure 3 This is the process of a robust invisible text watermark embedding method according to an embodiment of the present invention. Figure 3 ; Figure 4 This is the process of a robust invisible text watermark extraction method according to an embodiment of the present invention. Figure 1 ; Figure 5 This is the process of a robust invisible text watermark extraction method according to an embodiment of the present invention. Figure 2 ; Figure 6 This is the process of a robust invisible text watermark extraction method according to an embodiment of the present invention. Figure 3 . DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0023] It should be noted that the illustrations provided in the embodiments of the present invention are only schematic illustrations of the basic concept of the present invention. Therefore, the drawings only show components related to the present invention and are not drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component can be changed at will, and the component layout type may also be more complicated.

[0024] In the present invention, it should also be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer" and the like are used to indicate positions or locations based on those shown in the accompanying drawings. These terms are intended solely to facilitate the description of the present application and to simplify the description. They are not intended to indicate or imply that the devices or components referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present application. Furthermore, the terms "first" and "second" are used solely for descriptive and distinguishing purposes and should not be construed as indicating or implying relative importance.

[0025] In addition, it should be noted that the features of the various embodiments of the present invention may be combined or coupled in part or in whole, and, as will be appreciated by those skilled in the art, may interact and operate in different ways. Each embodiment may be implemented independently of one another or in an associated relationship.

[0026] The present invention mainly embeds invisible watermarks into short plain texts based on the characteristics of Chinese characters, thereby improving the robustness of watermark embedding.

[0027] Example 1 like Figure 1 As shown, the robust invisible text watermark embedding method of this embodiment mainly includes the following steps: S101, segmenting the text carrier to be watermarked, with each field corresponding to a bit of the bit string to be embedded in the watermark; S102, determine whether each bit of the bit string embedded with the watermark is 1, if so, execute step S103, if not, execute step S104; S103. If the bit is 1, the word at the specified position of the field corresponding to the bit is modified to a word whose first letter is the letter before the preset letter, and the semantics of the field remains unchanged after the modification; S104. If the bit is 0, the word at the specified position of the field corresponding to the bit is modified to a word whose first letter is not the letter before the preset letter, and the semantics of the field remains unchanged after the modification.

[0028] The designated position may be all the characters in the field, or the Chinese characters in the first half or the second half of the sub-segment, or the first character or the last character of the field.

[0029] Among them, the text carrier is segmented specifically as follows: the text carrier is evenly divided into multiple segments according to the number of bit strings embedded with the watermark; or, the text carrier is segmented according to the punctuation marks, and the method of selecting the segmentation is to judge based on the segmentation of the punctuation marks, that is, the sentence before each punctuation mark is regarded as a field, and the number of Chinese characters in the field is greater than L (for example, L=3).

[0030] In the specific implementation, we can first calculate the relevant parameters: (1) Calculate the total number of Chinese characters in the text carrier to be embedded with the watermark, assuming it is n. (2) Calculate the number of bits in the embedded watermark bit string, assuming it is m. Calculate n / m and round it down to the integer, which represents the number of Chinese characters required to embed 1 bit, assuming it is t.

[0031] The method for selecting the location for watermark embedding is as follows: (1) Take t Chinese characters and embed the watermark into the first half of the t characters. Specifically, if t is an odd number, the first half of the characters are from the 1st to the (t+1) / 2th characters. If t is an even number, the first half of the characters are from the 1st to the t / 2th characters.

[0032] (2) Embed the watermark into the second half of t Chinese characters. Specifically, if t is an odd number, the second half of the Chinese characters is from the (t+1) / 2+1th Chinese character to the tth Chinese character. If t is an even number, the second half of the Chinese characters is from the t / 2+1th Chinese character to the tth Chinese character.

[0033] (3) Embed the watermark into the middle part of t Chinese characters. Specifically, take the Chinese characters between the lower integer of t / 3 and the upper integer of (t / 3)·2.

[0034] The watermark embedding method or location can be agreed upon in advance, such as conveying the embedding location as a kind of information to the embedding party and the extraction party.

[0035] When embedding a watermark, it is usually necessary to copy the embedded bit string z times and embed it into the text repeatedly to improve the robustness of subsequent extraction.

[0036] Embed watermark w in sequence. Assume that the length of w is L, denote w as w[i]. Assume w="1010", then w[1]=1, w[2]=0, w[3]=1, w[4]=0.

[0037] Embed w[1], w[2], w[3], w[4] in sequence.

[0038] Take w[i] in sequence. If w[i] = 1, that is, embed 1: If the first letters of the pinyin of all Chinese characters in the Chinese character string x are all letters before m (m is called the segmentation letter, m is a system parameter, and it can also be other letters, such as n, such as o, as long as it is agreed in advance), then it remains unchanged; otherwise, replace this Chinese character string so that the first letters of the pinyin of this Chinese character string are all letters before m. Note that the meaning of the Chinese character string after replacement is the same as that before replacement.

[0039] If w[i] = 0, that is, embed 0: If the first letters of the pinyin of all Chinese characters in the Chinese character string x are not all letters before m, then it remains unchanged; otherwise, replace this Chinese character string so that the first letters of the pinyin of this Chinese character string are not all letters before m. Note that the meaning of the Chinese character string after replacement is close to that before replacement. Note that here m is a system parameter (because nearly half of the letters are before m and half are after m, it can be changed to n, or o, or other letters agreed in advance), which can be agreed in advance between the embedding party and the extraction party. For example, it can also be other letters. Note that whether it is the first letter or the last letter here is also a system parameter, which can be agreed in advance between the embedding party and the extraction party.

[0040] Furthermore, it can be further extended to select the last letter of the pinyin letters of the Chinese character string x for judgment, and the rest is the same as the above method.

[0041] Furthermore, no matter which modification method is used, if the original field needs to be modified to a field with the same semantics, a synonym replacement character library needs to be constructed to achieve this.

[0042] Among them, the construction method of the synonym replacement character library is: classify the synonyms according to the first letter of the pinyin, that is, either all the first letters of all Chinese characters in the two synonyms are before m, or not all are before m (assuming the agreed letter is m). Example: "If" and "For example" form a pair of replacement words. RG is not all before m, and BF is all before m. Construct the replacement character library according to the above method.

[0043] It should be noted that the establishment of the synonym character library is related to parameters. For example, if the segmentation letter is m, the method can select the best embedding position and method according to the actual situation of the text. It is necessary to scan the text multiple times to select the most suitable embedding position and embedding method, and then inform the extraction party of the parameters. Note that the judgment methods include: inputting the scheme after embedding the watermark into a pre-agreed natural language processing (NLP) model for evaluation, obtaining the return value (evaluation value), comparing the return values, and taking the parameter with the highest return value.

[0044] The embedding method can also be to judge which method is the best for the text after embedding in pinyin mode, the text after embedding in stroke mode, or the text after embedding in the odd or even number of characters mode. The judgment method is similar to the above.

[0045] Example 2 The main innovative ideas of this embodiment are similar to those of embodiment 1, except that the bits are embedded by modifying the number of strokes of the character at the specified position of the field to an odd number or an even number.

[0046] like Figure 2 As shown, the robust invisible text watermark embedding method of this embodiment mainly includes the following steps: S201, segmenting the text carrier to be embedded with the watermark, with each field corresponding to a bit of the bit string embedded with the watermark; S202, determine whether each bit of the bit string embedded with the watermark is 1, if so, execute step S203, if not, execute step S204; S203: If the bit is 1, the word at the specified position of the field corresponding to the bit is modified to a word with an odd or even number of strokes, and the semantics of the field remains unchanged after the modification; S204. If the bit is 0, the word at the specified position of the field corresponding to the bit is modified to a word with an even or odd number of strokes, which is exactly the opposite of the modification when the bit is 1. The semantics of the field remains unchanged after the modification.

[0047] In a specific embodiment, the parity of the number of strokes of the Chinese characters in the Chinese character string x is selected for judgment, that is, the strokes of all the Chinese characters in the Chinese character string x are all odd numbers (or all even numbers).

[0048] Embedding 1: If all characters in the string x have an odd (or even) number of strokes, leave it unchanged. Otherwise, replace the string so that all characters in the string have an odd (or even) number of strokes. Note that the replaced string has a similar meaning to the original string.

[0049] Embed 0: If the number of strokes of all characters in the string x is not all odd (or even), then leave it unchanged; otherwise, replace the string so that all characters in the string do not all have an odd (or even) number of strokes. Note that the replaced string has a similar meaning to the original string.

[0050] The extraction method is similar to the embedding method, which is to extract 1 based on whether the first letter (or last letter) of the Chinese character string x embedded with the watermark in t Chinese characters is before m (or other agreed letters). If so, extract 0, otherwise extract 0.

[0051] Among them, the construction of the synonym library is as follows: Synonyms are classified according to the parity of the number of Chinese characters' strokes, that is, the number of Chinese characters' strokes of two synonyms is either all odd or not all odd. Example: "If" and "假如", the former has 7 and 9 Chinese characters' strokes, all odd, and the latter has 12 and 9 Chinese characters' strokes, not all odd.

[0052] Embodiment 3 The main innovative idea of this embodiment is similar to that of Embodiment 1. The difference is that bit positions are embedded by modifying the total number of Chinese characters in the field to be odd or even.

[0053] As Figure 3 shown, the robust invisible text watermark embedding method of this embodiment mainly includes the following steps: S301. Segment the text carrier to be embedded with watermark, and each field corresponds to a bit position of the bit string to be embedded with watermark in turn; S302. Judge whether each bit position of the bit string to be embedded with watermark is 1. If so, execute step S203; if not, execute step S204; S303. If the bit position is 1, modify the number of Chinese characters in the field corresponding to this bit position to be odd, and the semantics of the field remain unchanged after modification S304. If the bit position is 0, modify the number of Chinese characters in the field corresponding to this bit position to be even, and the semantics of the field remain unchanged after modification.

[0054] Specifically, for embedding 1: If the number of Chinese characters in the Chinese character string x is odd, it remains unchanged; otherwise, replace this Chinese character string so that the number of Chinese characters in this Chinese character string is odd. Note that the Chinese character string after replacement and the Chinese character string before replacement have similar meanings. For embedding 0: If the number of Chinese characters in the Chinese character string x is even, it remains unchanged; otherwise, replace this Chinese character string so that the number of Chinese characters in this Chinese character string is even. Note that the Chinese character string after replacement and the Chinese character string before replacement have similar meanings. The adding method is: remove "的" in the Chinese character string x, add "即" in front of the Chinese character string x, add function words such as "吧", "啊" at the end of x, and add "了" after the verb.

[0055] The above embodiments give multiple watermark embedding methods. During specific implementation, one can be selected, or all can be used, and then compare the meanings of the fields after different embedding methods, and select a watermark embedding method with the closest field meaning. Specifically, it can be evaluated by a natural language processing (NLP) model which field meaning is closest to the original field meaning.

[0056] For example: A certain doctor uploads the patient's condition report text to the network, for example: Monitor the physical measurement indicators and laboratory test indicators of the child, and supplement the nutrition required for the child's growth. After nearly 3 months of care, the child's weight increased from 8.0 kg during the myelosuppression period after chemotherapy to 10.3 kg, and the hemoglobin concentration increased from 85 g / L to 114 g / L.

[0057] If the embedded watermark bit string C is "10010", which is the doctor's work number, C[1]=1; C[2]=0; C[3]=0; C[4]=1; C[5]=0.

[0058] If the fragment is cut according to punctuation as x, then x[1] = "Monitor the physical measurement indicators of the child". If C[1]=1 is embedded, the sentence remains unchanged. If C[1]=0 is embedded, then remove "of".

[0059] x[2] = "Laboratory test indicators". If C[2]=1 is embedded, the sentence remains unchanged. If C[2]=0 is embedded, the sentence is changed to: "Laboratory test indicators", making the number of characters even. That is, add "of" after the noun.

[0060] x[3] = "Supplement the nutrition required for the child's growth". If C[3]=0 is embedded, the sentence remains unchanged. If C[3]=1 is embedded, the sentence is changed to: "Supplement the nutrition required for the child's growth", that is, add "the" after the verb.

[0061] x[4] = "After nearly 3 months of care". If C[4]=0 is embedded, the sentence is changed to "After nearly 3 months of care", that is, remove "of". If C[4]=1 is embedded, the sentence is changed to: After nearly 3 months of care, making the number of characters even.

[0062] x[5] = "The child's weight increased from 8.0 kg during the myelosuppression period after chemotherapy to 10.3 kg". If C[5]=1 is embedded, the sentence remains unchanged. If C[5]=0 is embedded, the sentence becomes "That is, the child's weight increased from 8.0 kg during the myelosuppression period after chemotherapy to 10.3 kg". That is, add "That is" at the beginning of the sentence.

[0063] Example 4 This example is the watermark extraction method for embedding corresponding to Example 1.

[0064] As Figure 4 shown, this robust invisible text watermark extraction method includes the following steps: S401. Segment the text carrier, and each field corresponds to one bit position of the embedded watermark bit string in sequence; S402. Judge in sequence whether the first letter of the word at the specified position of each field is the letter before the preset letter. If so, execute step S403; if not, execute step S404; S403, the bit embedded in this field is 1; S404: The embedded bit of this field is 0.

[0065] Example 5 This embodiment is an embedded watermark extraction method corresponding to embodiment 1.

[0066] like Figure 5 As shown, the robust invisible text watermark extraction method includes the following steps: S501, segmenting the text carrier, with each field corresponding to a bit of the bit string embedded with the watermark; S502, determine in turn whether the number of strokes of the Chinese character at the specified position of each field is an odd number, if so, execute step S503; if not, execute step S504; S503, the bit embedded in this field is 1; S504: The embedded bit of this field is 0.

[0067] Example 5 This embodiment is an embedded watermark extraction method corresponding to embodiment 1.

[0068] like Figure 6 As shown, the robust invisible text watermark extraction method includes the following steps: S601, segmenting the text carrier, with each field corresponding to a bit of the bit string embedded with the watermark; S602, determine in turn whether the number of Chinese characters in each field is an odd number, if so, execute step S603; if not, execute step S604; S603, the bit embedded in this field is 1; S604: The embedded bit of this field is 0.

[0069] This application also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a disk, an optical disk, a server, an App store, etc., which stores a computer program that implements corresponding functions when executed by a processor. The computer-readable storage medium of this embodiment, when executed by a processor, implements the robust invisible text watermark embedding or extraction method of the method embodiment.

[0070] In summary, this invention can embed relatively short text (e.g., less than 100 Chinese characters) and is suitable for plain text, such as text in PDF files, Word documents, or other formatted text, as used in many existing solutions. This invisible watermark, designed based on the characteristics of Chinese characters, embeds only the text and characters without changing the original Chinese meaning of the field, making it a very robust watermark embedding method. Similarly, the watermark extraction method is also robust, resistant to text deletion and sentence permutation. Even if the text is partially deleted or sentences are rearranged, the watermark information can be largely recovered.

[0071] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, or two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0072] The size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0073] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A robust invisible text watermark embedding method, characterized by: The following steps are involved: S1. Segment the text carrier to be embedded with the watermark, and each field corresponds to a bit of the bit string embedded with the watermark; S2. Determine whether each bit of the bit string embedded with the watermark is 1 or 0. If it is 1, modify the word at the specified position of the field corresponding to the bit to a word whose first letter is the letter before the preset letter, and the semantics of the field remains unchanged after the modification; if it is 0, modify the word at the specified position of the field corresponding to the bit to a word whose first letter is not the letter before the preset letter, and the semantics of the field remains unchanged after the modification.

2. The robust invisible text watermark embedding method according to claim 1, characterized in that Alternatively, step S2 is specifically as follows: determining whether each bit of the bit string embedded with the watermark is 1 or 0; if it is 1, modifying the word at the specified position of the field corresponding to the bit to a word with an odd or even number of strokes, and the semantics of the field remain unchanged after the modification; if it is 0, modifying the word at the specified position of the field corresponding to the bit to a word with an even or odd number of strokes, and the modification is exactly opposite to the modification when the bit is 1, and the semantics of the field remain unchanged after the modification.

3. The robust invisible text watermark embedding method according to claim 1, characterized in that Alternatively, step S2 is specifically as follows: determining whether each bit of the bit string embedded with the watermark is 1 or 0; if it is 1, modifying the number of Chinese characters in the field corresponding to the bit to an odd number or an even number, and the semantics of the field remain unchanged after the modification; if it is 0, modifying the number of Chinese characters in the field corresponding to the bit to an even number or an odd number, and the modification is exactly opposite to the modification when the bit is 1, and the semantics of the field remain unchanged after the modification.

4. The robust invisible text watermark embedding method according to claim 3, characterized in that: When changing the number of Chinese characters to an odd number or an even number, the modification method includes removing or adding function words.

5. The robust invisible text watermark embedding method according to claim 1 or 2, characterized in that: The specified position is all the characters in the field, or the Chinese characters in the first half or the second half of the sub-segment, or the first or last character of the field.

6. The robust invisible text watermark embedding method according to claim 1 or 2, characterized in that: The text carrier is segmented specifically as follows: the text carrier is evenly divided into multiple segments according to the number of bit strings embedded with the watermark.

7. The robust invisible text watermark embedding method according to any one of claims 1 to 4, characterized in that: Segmenting the text carrier specifically includes segmenting the text carrier according to punctuation marks.

8. A robust invisible text watermark extraction method, characterized in that The following steps are involved: S1. Segment the text carrier, and each field corresponds to a bit of the bit string embedded in the watermark; S2. Determine in turn whether the first letter of the word at the specified position of each field is the letter before the preset letter. If so, the bit embedded in the field is 1; if not, the bit embedded in the field is 0.

9. The robust invisible text watermark extraction method according to claim 8, characterized in that: Alternatively, step S2 is specifically as follows: determining in turn whether the number of Chinese characters in each field is odd or even; if it is odd, the bit embedded in the field is 1; if it is even, the bit embedded in the field is 0.

10. The robust invisible text watermark extraction method according to claim 8, characterized in that: Alternatively, step S2 is specifically as follows: determining in turn whether the number of Chinese character strokes at the designated position of each field is odd or even; if it is odd, the bit embedded in the field is 1; if it is even, the bit embedded in the field is 0.

Images