Voice processing method and device

Abstract

The present invention discloses a speech processing method and device. The invention comprises: obtaining target speech data; cutting the target speech data into multiple sub-segments of speech data; identifying noise data contained in the multiple sub-segments of speech data, and removing the identified noise data. The present invention solves the technical problem of low efficiency of noise identification methods in related technologies.

Description

Technical Field

[0001] The present invention relates to the field of speech processing, and in particular to a speech processing method and device. Background Art

[0002] In related technologies, as speech recognition is increasingly used in life, various fields attach more and more importance to the results of speech recognition. The results of speech recognition are not only affected by the language model, but also closely related to the quality of speech. However, in actual production, the recorded audio may contain a lot of noise due to interference from the surrounding environment, so the removal of noise data comes into being.

[0003] Among the existing noise recognition methods, there is a method for removing noise in audio by using the difference between short-time energy and average short-time energy. This method can effectively remove noise in audio signals. However, the noise removal technology in the existing technology may cause misjudgment, resulting in low accuracy in noise judgment.

[0004] With respect to the above-mentioned problems existing in the related technologies, no effective solution has been proposed yet. Summary of the invention

[0005] The main purpose of the present invention is to provide a speech processing method and device to solve the technical problem of low efficiency of noise recognition methods in related technologies.

[0006] In order to achieve the above object, according to one aspect of the present invention, a method for processing speech is provided, which comprises: obtaining target speech data; cutting the target speech data into multiple sub-segments of speech data; identifying noise data contained in the multiple sub-segments of speech data, and removing the identified noise data.

[0007] Furthermore, obtaining the target voice data includes: obtaining a target voice segment, and determining information of the target voice segment based on the target voice segment, wherein the information of the target voice segment includes at least the number of sampling bits of the target voice segment and the sampling frequency of the target voice segment; judging whether the information of the target voice segment meets preset requirements; if the information of the target voice segment meets the preset requirements, converting the target voice segment into target voice data through analog-to-digital conversion.

[0008] Furthermore, cutting the target voice data into multiple sub-segments of voice data includes: utilizing silence segments contained in the target voice data to cut the target voice data into multiple sub-segments of voice data, wherein the time of the silence segment is a preset time, and the silence segment is a segment that does not contain valid sound.

[0009] Furthermore, after the target voice data is cut into multiple segments of sub-voice data, the method includes: marking the sub-voice data to obtain the corresponding normal voice time period and noise time period in the sub-voice data; obtaining the audio value distribution in the noise time period by preset means; and obtaining the noise value and the range of the noise value contained in the sub-voice data by the audio value distribution.

[0010] Furthermore, identifying noise data contained in multiple segments of sub-voice data includes: reading multiple audio values ​​contained in the sub-voice data, and determining the cumulative value of the multiple audio values; judging whether the cumulative value is greater than a threshold value; if the cumulative value is greater than the threshold value, determining the sub-voice data as noise data.

[0011] Furthermore, identifying noise data contained in multiple segments of sub-voice data also includes: identifying the sub-voice data corresponding to the normal voice time period and obtaining a recognition result, wherein the recognition result includes at least the number of characters included in the sub-voice data and the duration of the voice segment corresponding to the sub-voice data; determining the frequency of characters in the sub-voice data corresponding to the normal voice time period based on the number of characters and the duration of the voice segment; subtracting the character frequency from the preset character frequency; if the difference between the character frequency and the preset character frequency is not within the preset range, determining the sub-voice data corresponding to the normal voice time period as noise data.

[0012] In order to achieve the above object, according to another aspect of the present invention, a speech processing device is provided. The device comprises: a first acquisition unit, used to acquire target speech data; a cutting unit, used to cut the target speech data into multiple sub-segments of speech data; and a recognition unit, used to recognize noise data contained in the multiple sub-segments of speech data and remove the recognized noise data.

[0013] In order to achieve the above object, according to another aspect of the present invention, a computer-readable storage medium is provided, which includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned speech processing method.

[0014] In order to achieve the above object, according to another aspect of the present invention, a processor is provided, the processor being used to run a program, wherein the program executes the above speech processing method when it is run.

[0015] Through the present invention, the following steps are adopted: obtaining target voice data; cutting the target voice data into multiple sub-segments of voice data; identifying the noise data contained in the multiple sub-segments of voice data, and eliminating the identified noise data, thereby solving the technical problem of low efficiency of the noise identification method in the related art, and further reducing the impact of noise data on actual production. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0017] Figure 1 is a flow chart of a method for processing speech according to an embodiment of the present invention;

[0018] Figure 2 The method for processing speech provided by the embodiment of the present invention is Figure 1 Specific flow diagram of step S101;

[0019] Figure 3 The method for processing speech provided by the embodiment of the present invention is Figure 1 Specific flow diagram of step S102;

[0020] Figure 4 The method for processing speech provided by the embodiment of the present invention is Figure 1 Another flow chart after step S102;

[0021] Figure 5 The method for processing speech provided by the embodiment of the present invention is Figure 1 Specific flow diagram of step S103;

[0022] Figure 6 The method for processing speech provided by the embodiment of the present invention is Figure 1 Another flow chart of step S103;

[0023] Figure 7 It is a schematic diagram of a speech processing device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present invention described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] Example 1

[0028] According to an embodiment of the present invention, a method for processing speech is provided.

[0029] According to an embodiment of the present invention, a method embodiment of a speech processing method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0030] Figure 1 1 is a flow chart of a method for processing speech according to an embodiment of the present invention. The method can be applied to a terminal device or a server, wherein the terminal device can be a computing device such as a desktop computer, a notebook, a handheld computer, etc., and the server can be a computing device such as a cloud server, etc. Figure 1 As shown, the speech processing method includes the following steps:

[0031] Step S101, obtaining target voice data.

[0032] Optionally, the target voice data here may be a segment of voice uttered by a predetermined object.

[0033] In this embodiment, when applied to a terminal device, a microphone or a microphone array can be used to obtain the voice of a predetermined object; when applied to a server, the voice of the predetermined object can be obtained from a peripheral device of the server (for example, a call center).

[0034] In this embodiment, it is first necessary to obtain a segment of speech data to be processed (ie, target speech data), wherein the speech data may be speech data in any scenario.

[0035] Optionally, the above-mentioned voice data has no limitation on any specific requirements such as voice type and voice length, and can be any type of voice; in addition, in the embodiment of the present invention, there is no specific limitation on the source of voice acquisition.

[0036] In an alternative embodiment, see Figure 2 , a method for processing speech provided by an embodiment of the present invention Figure 1 The specific flow chart of step S101 in the embodiment of the present invention is to obtain the target voice data, including:

[0037] Step S201, obtaining a target speech segment, and determining information of the target speech segment according to the target speech segment, wherein the information of the target speech segment at least includes the number of sampling bits of the target speech segment and the sampling frequency of the target speech segment;

[0038] First, you need to input a speech segment and obtain basic information such as the sampling bit number and sampling frequency of the input speech segment.

[0039] Among them, the above-mentioned sampling frequency refers to the number of times the analog signal is sampled per unit time. The higher the sampling frequency, the more realistic the sound restoration effect, and of course the more data is sampled; the number of sampling bits refers to the data range that each sampling point can represent. The larger the number of sampling bits, the more delicate the sound.

[0040] Step S202, determining whether the information of the target voice segment meets the preset requirements;

[0041] It can be understood that, in this embodiment, it is necessary to determine whether the voice segment meets the requirements of the preset sampling frequency and the preset sampling bit number.

[0042] Step S203: If the information of the target voice segment meets the preset requirements, the target voice segment is converted into target voice data through analog-to-digital conversion.

[0043] Optionally, if the voice segment does not meet the preset sampling frequency and the preset sampling position, another voice segment needs to be re-imported. It should be noted that if the first imported voice segment meets the preset requirements, the first imported voice segment is converted into target voice data, that is, the voice segment to be processed is converted into target voice data through analog-to-digital conversion.

[0044] As described above, in this embodiment, the preset requirements include at least a preset sampling frequency and a preset sampling position. At the same time, in different scenarios and requirements, the voice segment to be processed may also include other preset requirements, which are not specifically limited in this embodiment.

[0045] It should be further explained that in real life, the time of the sound clips we can hear is continuous. The above-mentioned voice clips are analog signals, but when the audio data is stored and processed in the computer, the analog signal needs to be converted into a digital signal. Therefore, the voice clips to be processed that meet the requirements need to be converted into target voice data through analog-to-digital conversion.

[0046] Furthermore, if the first voice segment input does not meet the preset requirements, it is necessary to re-input another voice segment, and after inputting another voice segment, the above steps S201 to S203 are repeated.

[0047] Step S102, cutting the target voice data into multiple segments of sub-voice data.

[0048] In an alternative embodiment, see Figure 3 , Figure 3 The method for processing speech provided by the embodiment of the present invention is Figure 1 The specific flow chart of step S102 in the embodiment of the present invention is to cut the target voice data into multiple sub-voice data segments, including:

[0049] Step S301, using the silence segment contained in the target voice data, the target voice data is cut into multiple sub-segments of voice data, wherein the time of the silence segment is a preset time, and the silence segment is a segment that does not contain valid sound.

[0050] In the above, using the silence segment contained in the target voice data to cut the target voice data into multiple segments of sub-voice data includes the following specific steps:

[0051] Step 1: Determine the duration of the silent segment, wherein the so-called silent segment is a segment during which there is no sound or no one speaks, that is, a segment without the required effective sound is a silent segment;

[0052] It should be noted that the silence segment time should not be set too long. If the time is too long, the data after audio cutting may still contain a large amount of silence time, and the noise data cannot be cut out; if the silence segment time is too short, the audio cutting will be more fragmented, and the audio recognition effect will be reduced during subsequent work. Preferably, the silence segment time is set to 500ms.

[0053] Step 2: Use the silence segments to cut the target voice data into several segments of voice data.

[0054] It should also be noted that the target voice data is cut into multiple sub-voice data through the segments included in the target voice data. When the noise data is recognized later, the silent segments that do not contain valid sounds are cut off in advance, thereby improving the efficiency of noise recognition.

[0055] In an alternative embodiment, see Figure 4 , a method for processing speech provided by an embodiment of the present invention Figure 1 Another flow chart after step S102 in FIG. 1 , after cutting the target voice data into multiple segments of sub-voice data, the method includes:

[0056] Step S401: annotate the sub-speech data to obtain the corresponding normal speech time period and noise time period in the sub-speech data.

[0057] Optionally, the sub-voice data can be labeled manually or by machine learning. After the sub-voice data is labeled, a labeled audio will be generated accordingly. By analyzing the labeled time period through the labeled audio, the normal voice time period corresponding to the normal voice information and the noise time period containing the noise fragments in the sub-voice data can be obtained.

[0058] Step S402: obtaining the distribution of audio values ​​in the noise time period by a preset means.

[0059] Specifically, after the corresponding noise time period is determined by marking the sub-speech data, the speech data corresponding to the noise time period needs to be processed, that is, the audio value distribution in the noise time period is obtained through preset means, wherein the audio value distribution in the noise time period can be displayed through a curve graph or through numerical representation.

[0060] Among them, preferably, the preset means is a voting mechanism, which is to use the audio values ​​that appear in the noise time period to achieve accumulation. For example, if in the noise time period, one of the audio values ​​is 100, and the number of existing audio values ​​100 is 50 times, when 100 appears again, the number of times the value 100 appears becomes 51.

[0061] It should be noted that in addition to the above-mentioned voting mechanism, other technical means can also be used to determine the distribution of audio values ​​during the noise period.

[0062] Step S403, obtaining the noise value and the range of the noise value contained in the sub-speech data according to the distribution of the audio value.

[0063] At least the noise value and the noise value range contained in the sub-speech data can be determined by the distribution of the audio value in the noise time period.

[0064] Therefore, by obtaining the distribution of audio values ​​in the noise time period as described above, the sound conditions in the noise time period can be intuitively obtained, and a numerical basis for determining the noise can be obtained.

[0065] Step S103, identifying the noise data contained in the multiple segments of sub-voice data, and removing the identified noise data.

[0066] In an alternative embodiment, see Figure 5 , a method for processing speech provided by an embodiment of the present invention Figure 1 The specific flow chart of step S103 in the embodiment of the present invention is to identify the noise data contained in the multiple segments of sub-speech data, including:

[0067] Step S501, reading a plurality of audio values ​​contained in the sub-voice data, and determining a cumulative value of the plurality of audio values.

[0068] In the above, after obtaining the noise value and the noise value range in the sub-speech data, the value corresponding to the noise data can be determined by the audio value and the audio value range included in the marked noise time period.

[0069] The secondary determination of the sub-voice data further includes reading a plurality of audio values ​​contained in the sub-voice data, and accumulating the plurality of audio values ​​to obtain an accumulated value.

[0070] Step S502, determining whether the accumulated value is greater than a threshold value.

[0071] Step S503: if the accumulated value is greater than the threshold value, the sub-speech data is determined to be noise data.

[0072] Specifically, in an optional embodiment, if the audio values ​​in the sub-voice data are all within the range of noise values, the audio values ​​in the sub-voice data are accumulated to obtain a cumulative value. If the cumulative value is greater than a threshold value, it indicates that the sub-voice data is noise data.

[0073] This embodiment also provides another optional embodiment, in which it is determined whether the audio values ​​contained in the read sub-voice data include audio values ​​that belong to a noise time period. If there are audio values ​​that belong to the noise time period, the corresponding data at that time point is determined to be noise data.

[0074] Since a narrow range of noise values ​​determined by the above steps may easily lead to misjudgment of information, a secondary judgment is performed on the sub-speech data through the above steps S501-S503.

[0075] Through the above two optional embodiments, noise data can be accurately determined, avoiding the problem of mistaking noise data for normal voice data due to individual audio values ​​being too high or too low, and also avoiding the technical problem of low accuracy in determining noise data due to a narrow range of noise values.

[0076] In another alternative embodiment, see Figure 6 , a method for processing speech provided by an embodiment of the present invention Figure 1 Another flow chart of step S103 in the embodiment of the present invention is to identify the noise data contained in the multiple segments of sub-speech data, including:

[0077] Step S601, recognize the sub-voice data corresponding to the normal voice time period and obtain a recognition result, wherein the recognition result at least includes the number of characters included in the sub-voice data and the duration of the voice segment corresponding to the sub-voice data.

[0078] Step S602, determining the frequency of the characters in the sub-speech data corresponding to the normal speech time period according to the number of characters and the duration of the speech segment.

[0079] Step S603, subtracting the text frequency from the preset text frequency.

[0080] Step S604: if the difference between the text frequency and the preset text frequency is not within the preset range, the sub-speech data corresponding to the normal speech time period is determined as noise data.

[0081] As described above, after determining the audio value in the noise time period, it is necessary to further determine the audio value in the normal speech time period to further accurately determine the noise data to avoid omissions that may cause low determination accuracy.

[0082] By determining the text frequency of the sub-speech data corresponding to the normal speech time period, the determined text frequency is subtracted from the preset text frequency. When the difference between the two values ​​is large (that is, when the difference between the two is not within the preset range), the data in the normal speech time period is determined as noise data, otherwise it is normal audio data.

[0083] For example, the speaking speed calculated in the business is 60 words / s, but the noise data may be 80 words / s. In such cases, it can be judged as noise data.

[0084] Among them, the preset text frequency is pre-set for different scenarios. For example, the normal speaking speed is about 280 words (including punctuation) / min, and the normal speaking speed is about 46 words / s. However, in the telemarketing scenario, the salesperson needs to let the customer know the product information as soon as possible, so the speaking speed is too fast and the number of words spoken per second is relatively large; in the service scenario, the service staff must give the customer a good experience, so the speaking speed will be slower.

[0085] Therefore, the above steps can solve the problem of different speaking speeds in different scenarios.

[0086] Furthermore, the determined noise data is cut off and normal audio data is returned.

[0087] In an embodiment of the present invention, noise information can be judged for audio data in different scenarios. First, the audio information is segmented using silence, and then the noise data is removed from the segmented data. A voting mechanism is used to obtain the noise data range in the current scenario. When the noise range is set to be narrow, it will cause misjudgment of the information. Therefore, the audio is judged a second time based on the principle that there is a difference in the number of characters recognized by normal signals and noise signals in speech recognition, so as to further accurately screen out the noise data, thereby improving the preprocessing results of the audio information and reducing the impact of noise data on actual production.

[0088] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0089] Example 2

[0090] The embodiment of the present invention further provides a speech processing device. It should be noted that the Z device of the embodiment of the present invention can be used to execute a speech processing method provided by the embodiment of the present invention. The speech processing device provided by the embodiment of the present invention is introduced below.

[0091] Figure 7 is a schematic diagram of a speech processing device provided according to an embodiment of the present invention. Figure 7 As shown, the device includes: a first acquisition unit 701, used to acquire target voice data; a cutting unit 702, used to cut the target voice data into multiple segments of sub-voice data; and a recognition unit 703, used to recognize the noise data contained in the multiple segments of sub-voice data and remove the recognized noise data.

[0092] A speech processing device provided in an embodiment of the present invention is used to obtain target speech data through a first acquisition unit 701; a cutting unit 702 is used to cut the target speech data into multiple sub-segments of speech data; and a recognition unit 703 is used to recognize noise data contained in the multiple sub-segments of speech data and eliminate the recognized noise data, thereby solving the technical problem of low efficiency of noise recognition methods in related technologies, thereby reducing the impact of noise data on actual production.

[0093] Optionally, the first acquisition unit includes: an acquisition subunit, used to acquire a target speech segment, and determine information of the target speech segment based on the target speech segment, wherein the information of the target speech segment includes at least the number of sampling bits of the target speech segment and the sampling frequency of the target speech segment; a first judgment subunit, used to judge whether the information of the target speech segment meets preset requirements; and a conversion subunit, used to convert the target speech segment into target speech data through analog-to-digital conversion when the information of the target speech segment meets the preset requirements.

[0094] Optionally, the cutting unit 702 includes: a cutting subunit, used to use the silence segment contained in the target voice data to cut the target voice data into multiple sub-segments of voice data, wherein the time of the silence segment is a preset time and the silence segment is a segment that does not contain valid sound.

[0095] Optionally, the device also includes: a labeling unit, which is used to label the sub-voice data after cutting the target voice data into multiple sub-voice data segments, so as to obtain the corresponding normal voice time period and noise time period in the sub-voice data; a second acquisition unit, which is used to obtain the audio value distribution in the noise time period by preset means; and a third acquisition unit, which is used to obtain the noise value and the range of the noise value contained in the sub-voice data through the audio value distribution.

[0096] Optionally, the recognition unit 703 includes: a first determination subunit, used to read multiple audio values ​​contained in the sub-voice data and determine the cumulative value of the multiple audio values; a second judgment subunit, used to judge whether the cumulative value is greater than a threshold value; and a second determination subunit, used to determine the sub-voice data as noise data when the cumulative value is greater than the threshold value.

[0097] Optionally, the recognition unit 703 also includes: a recognition subunit, which is used to recognize the sub-voice data corresponding to the normal voice time period and obtain a recognition result, wherein the recognition result at least includes the number of characters included in the sub-voice data and the duration of the voice segment corresponding to the sub-voice data; a third determination subunit, which is used to determine the frequency of characters in the sub-voice data corresponding to the normal voice time period based on the number of characters and the duration of the voice segment; a difference subunit, which is used to differ the character frequency from the preset character frequency; and a fourth determination subunit, which is used to determine the sub-voice data corresponding to the normal voice time period as noise data when the difference between the character frequency and the preset character frequency is not within a preset range.

[0098] A speech processing device includes a processor and a memory. The first acquisition unit 701 and the like are stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions.

[0099] The processor includes a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels can be set, and the technical problem of low efficiency of the noise recognition method in the related art can be solved by adjusting the kernel parameters.

[0100] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0101] Example 3

[0102] An embodiment of the present invention provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a method for processing speech is implemented.

[0103] Example 4

[0104] An embodiment of the present invention provides a processor, which is used to run a program, wherein a speech processing method is executed when the program is run.

[0105] An embodiment of the present invention provides a device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: acquiring target voice data; cutting the target voice data into multiple segments of sub-voice data; identifying noise data contained in the multiple segments of sub-voice data, and removing the identified noise data.

[0106] Optionally, obtaining the target voice data includes: obtaining a target voice segment, and determining information of the target voice segment based on the target voice segment, wherein the information of the target voice segment includes at least the number of sampling bits of the target voice segment and the sampling frequency of the target voice segment; judging whether the information of the target voice segment meets preset requirements; if the information of the target voice segment meets the preset requirements, converting the target voice segment through analog-to-digital conversion to obtain the target voice data.

[0107] Optionally, cutting the target voice data into multiple sub-segments of voice data includes: utilizing silence segments contained in the target voice data to cut the target voice data into multiple sub-segments of voice data, wherein the time of the silence segment is a preset time, and the silence segment is a segment that does not contain valid sound.

[0108] Optionally, after cutting the target voice data into multiple segments of sub-voice data, the method includes: marking the sub-voice data to obtain the corresponding normal voice time period and noise time period in the sub-voice data; obtaining the audio value distribution in the noise time period by preset means; and obtaining the noise value and the range of the noise value contained in the sub-voice data by the audio value distribution.

[0109] Optionally, identifying noise data contained in multiple segments of sub-voice data includes: reading multiple audio values ​​contained in the sub-voice data, and determining a cumulative value of the multiple audio values; judging whether the cumulative value is greater than a threshold value; if the cumulative value is greater than the threshold value, determining the sub-voice data as noise data.

[0110] Optionally, identifying noise data contained in multiple sub-segments of sub-speech data also includes: identifying the sub-speech data corresponding to the normal speech time period and obtaining a recognition result, wherein the recognition result includes at least the number of characters included in the sub-speech data and the duration of the speech segment corresponding to the sub-speech data; determining the frequency of characters in the sub-speech data corresponding to the normal speech time period based on the number of characters and the duration of the speech segment; subtracting the frequency of characters from a preset frequency of characters; if the difference between the frequency of characters and the preset frequency of characters is not within a preset range, determining the sub-speech data corresponding to the normal speech time period as noise data. The device in this article can be a server, PC, PAD, mobile phone, etc.

[0111] Example 4

[0112] The present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: obtaining target voice data; cutting the target voice data into multiple segments of sub-voice data; identifying noise data contained in the multiple segments of sub-voice data, and removing the identified noise data.

[0113] Optionally, obtaining the target voice data includes: obtaining a target voice segment, and determining information of the target voice segment based on the target voice segment, wherein the information of the target voice segment includes at least the number of sampling bits of the target voice segment and the sampling frequency of the target voice segment; judging whether the information of the target voice segment meets preset requirements; if the information of the target voice segment meets the preset requirements, converting the target voice segment through analog-to-digital conversion to obtain the target voice data.

[0114] Optionally, cutting the target voice data into multiple sub-segments of voice data includes: utilizing silence segments contained in the target voice data to cut the target voice data into multiple sub-segments of voice data, wherein the time of the silence segment is a preset time, and the silence segment is a segment that does not contain valid sound.

[0115] Optionally, after cutting the target voice data into multiple segments of sub-voice data, the method includes: marking the sub-voice data to obtain the corresponding normal voice time period and noise time period in the sub-voice data; obtaining the audio value distribution in the noise time period by preset means; and obtaining the noise value and the range of the noise value contained in the sub-voice data by the audio value distribution.

[0116] Optionally, identifying noise data contained in multiple segments of sub-voice data includes: reading multiple audio values ​​contained in the sub-voice data, and determining a cumulative value of the multiple audio values; judging whether the cumulative value is greater than a threshold value; if the cumulative value is greater than the threshold value, determining the sub-voice data as noise data.

[0117] Optionally, identifying noise data contained in multiple segments of sub-voice data also includes: identifying the sub-voice data corresponding to the normal voice time period and obtaining a recognition result, wherein the recognition result includes at least the number of characters included in the sub-voice data and the duration of the voice segment corresponding to the sub-voice data; determining the frequency of characters in the sub-voice data corresponding to the normal voice time period based on the number of characters and the duration of the voice segment; subtracting the character frequency from the preset character frequency; if the difference between the character frequency and the preset character frequency is not within a preset range, determining the sub-voice data corresponding to the normal voice time period as noise data.

[0118] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0119] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0120] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0122] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0123] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0124] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0125] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0126] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0127] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for processing speech, It is characterized in that include: Obtain target voice data; Cutting the target voice data into multiple segments of sub-voice data; Identify the noise data contained in the plurality of sub-speech data segments, and remove the identified noise data. After the target voice data is cut into a plurality of sub-segments of voice data, the method comprises: Annotating the sub-speech data to obtain a normal speech time period and a noise time period corresponding to the sub-speech data; Obtaining the distribution of audio values ​​within the noise time period by a preset means; The noise value and the range of the noise value contained in the sub-speech data are obtained through the distribution of the audio value. Identifying the noise data contained in the plurality of sub-speech data further includes: Recognize the sub-voice data corresponding to the normal voice time period and obtain a recognition result, wherein the recognition result at least includes the number of characters included in the sub-voice data and the duration of the voice segment corresponding to the sub-voice data; Determining the frequency of the characters in the sub-speech data corresponding to the normal speech time period according to the number of characters and the duration of the speech segment; Subtracting the text frequency from a preset text frequency; If the difference between the text frequency and the preset text frequency is not within the preset range, the sub-voice data corresponding to the normal voice time period is determined as the noise data.

2. The method according to claim 1, It is characterized in that Obtaining target voice data includes: Acquire a target speech segment, and determine information of the target speech segment according to the target speech segment, wherein the information of the target speech segment at least includes the number of sampling bits of the target speech segment and the sampling frequency of the target speech segment; Determining whether the information of the target voice segment meets the preset requirements; If the information of the target voice segment meets the preset requirement, the target voice segment is converted into the target voice data through analog-to-digital conversion.

3. The method according to claim 1, It is characterized in that Cutting the target voice data into multiple segments of sub-voice data includes: The target voice data is cut into a plurality of sub-voice data segments by using the silence segments contained in the target voice data, wherein the duration of the silence segments is a preset duration and the silence segments are segments that do not contain valid sounds.

4. The method according to claim 1, It is characterized in that Identifying the noise data contained in the plurality of sub-speech data includes: Reading a plurality of audio values ​​contained in the sub-voice data, and determining a cumulative value of the plurality of audio values; Determine whether the accumulated value is greater than a threshold value; If the accumulated value is greater than the threshold value, the sub-voice data is determined to be the noise data.

5. A speech processing device, It is characterized in that include: A first acquisition unit, used to acquire target voice data; A cutting unit, used for cutting the target voice data into multiple segments of sub-voice data; A recognition unit, used to recognize noise data contained in the plurality of sub-speech data segments, and remove the recognized noise data; A labeling unit, for labeling the sub-speech data after cutting the target speech data into a plurality of sub-speech data segments, so as to obtain the corresponding normal speech time segment and noise time segment in the sub-speech data; A second acquisition unit is used to obtain the audio value distribution in the noise time period by a preset means; The third acquisition unit is used to obtain the noise value and the range of the noise value contained in the sub-speech data according to the distribution of the audio value. The recognition unit includes: a recognition subunit, which is used to recognize the sub-voice data corresponding to the normal voice time period and obtain a recognition result, wherein the recognition result at least includes the number of characters included in the sub-voice data and the duration of the voice segment corresponding to the sub-voice data; a third determination subunit, which is used to determine the frequency of characters in the sub-voice data corresponding to the normal voice time period based on the number of characters and the duration of the voice segment; a difference subunit, which is used to subtract the character frequency from the preset character frequency; and a fourth determination subunit, which is used to determine the sub-voice data corresponding to the normal voice time period as noise data when the difference between the character frequency and the preset character frequency is not within a preset range.

6. The device according to claim 5, It is characterized in that The first acquiring unit includes: An acquisition subunit, configured to acquire a target speech segment and determine information of the target speech segment according to the target speech segment, wherein the information of the target speech segment at least includes the number of sampling bits of the target speech segment and the sampling frequency of the target speech segment; A first judging subunit, used to judge whether the information of the target voice segment meets a preset requirement; The conversion subunit is used to convert the target voice segment into the target voice data through analog-to-digital conversion when the information of the target voice segment meets the preset requirement.

7. A computer-readable storage medium, It is characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the speech processing method described in any one of claims 1 to 4.

8. A processor, It is characterized in that The processor is used to run a program, wherein the program executes a speech processing method as described in any one of claims 1 to 4 when running.

Images