Processing method and system based on streaming text speech synthesis and broadcast

Abstract

The invention discloses a processing method and system based on streaming text speech synthesis and broadcast, and the method comprises the steps: receiving a streaming text which needs speech synthesis and broadcast, and segmenting the streaming text according to punctuation marks and line feed marks to obtain a current frame text; performing voice synthesis on the current frame text to obtain audio data corresponding to the current frame text; performing audio decoding on the audio data to obtain decoded audio data, calculating the playing duration of the decoded audio data, and playing the decoded audio data; and according to the playing duration, setting time before the decoded audio data is played, and repeatedly performing streaming text segmentation, speech synthesis, audio decoding and audio playing processing until streaming text speech synthesis and broadcasting are completed. According to the method and the device, the uplink supports streaming text input, the downlink supports streaming audio data return, the voice broadcast delay is reduced, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of speech synthesis, and in particular to a processing method and system based on streaming text speech synthesis and broadcasting. Background Art

[0002] Current smart TVs have the ability to use large-model voice. The large model returns streaming text for input questions, and then voice broadcasts these streaming texts. Existing speech synthesis and broadcasting solutions are generally implemented through cloud-based speech synthesis engines. Downlink supports streaming return of audio data, but uplink does not support streaming input text. After receiving all the texts that need speech synthesis, the texts are sent to the background together. The background streams back the audio data after speech synthesis, and then the player plays the audio data.

[0003] Obviously, in the scenario where a large model returns streaming text in Markdown format, if you wait for all the text to be returned before initiating speech synthesis, the user will have to wait a long time to hear the played sound. Therefore, the existing speech synthesis and broadcasting solutions cannot complete the speech synthesis and broadcasting of streaming text well, and there are problems such as large delay, severe pause, and poor user experience. Summary of the invention

[0004] The main purpose of the present invention is to provide a processing method and system based on streaming text-to-speech synthesis and broadcasting, aiming to solve the problems of large delay and severe pause in existing streaming text-to-speech synthesis and broadcasting solutions.

[0005] To achieve the above-mentioned object of the invention, the present invention provides a processing method based on streaming text-to-speech synthesis and broadcasting, and the processing method based on streaming text-to-speech synthesis and broadcasting comprises:

[0006] Receiving a stream text that needs to be synthesized and broadcasted, and segmenting the stream text according to punctuation marks and line breaks to obtain a current frame text;

[0007] Performing speech synthesis on the current frame text to obtain audio data corresponding to the current frame text;

[0008] Performing audio decoding on the audio data to obtain decoded audio data, calculating a playback duration of the decoded audio data, and playing the decoded audio data;

[0009] According to the playing time, a time is set before the decoded audio data is played, and the streaming text segmentation, speech synthesis, audio decoding and audio playing processing are repeated until the streaming text speech synthesis and broadcasting are completed.

[0010] Optionally, the receiving of the streaming text that needs to be voice synthesized and broadcasted is specifically:

[0011] Receive the streaming text output by the large model through the client that needs to be synthesized and broadcasted;

[0012] The streaming text includes text in Markdown format.

[0013] Optionally, segmenting the streaming text according to punctuation marks and line breaks to obtain the current frame text specifically includes:

[0014] On the client, using punctuation marks and line breaks as identifiers, querying the received streaming text from back to front;

[0015] If the identifier is found, the identifier and the text before the identifier are segmented to obtain the current frame text;

[0016] The current frame text is sent to the cloud.

[0017] Optionally, performing speech synthesis on the current frame text to obtain audio data corresponding to the current frame text specifically includes:

[0018] In the cloud, using a TTS method to perform speech synthesis on the current frame text to obtain audio data corresponding to the current frame text;

[0019] Returning the audio data to the client;

[0020] Wherein, the audio data includes audio data in mp3 format.

[0021] Optionally, the performing audio decoding on the audio data to obtain decoded audio data, and calculating the playback duration of the decoded audio data specifically includes:

[0022] On the client, performing audio decoding on the audio data to obtain decoded audio data, wherein the decoded audio data includes audio data in a pcm format;

[0023] Calculate the playback duration of the decoded audio data:

[0024] T = L × 8 / (F × A × B);

[0025] Among them, T represents the playback time, L represents the size of the decoded audio data, F represents the sampling rate, A represents the number of sampling bits, and B represents the number of channels.

[0026] Optionally, playing the decoded audio data is specifically:

[0027] On the client, playing the decoded audio data through a player;

[0028] Among them, the different decoded audio data corresponding to texts with different frame numbers are played by a common player.

[0029] Optionally, the step of setting a time before playing the decoded audio data according to the playing time, repeating the streaming text segmentation, speech synthesis, audio decoding and audio playing processing until the streaming text speech synthesis and broadcasting are completed, specifically includes:

[0030] According to the playing time, a time is set before the decoded audio data is played, and the streaming text except the current frame text is segmented according to punctuation marks and line breaks to obtain the next frame text;

[0031] Performing speech synthesis, audio decoding and audio playback processing on the next frame of text;

[0032] The streaming text segmentation, speech synthesis, audio decoding and audio playing processes are repeated until the entire streaming text speech synthesis and broadcasting are completed.

[0033] To achieve the above-mentioned object of the invention, the present invention further provides a processing system based on streaming text-to-speech synthesis and broadcasting, the processing system based on streaming text-to-speech synthesis and broadcasting comprising:

[0034] Text processing module: used for receiving streaming text that needs to be synthesized and broadcast, segmenting the streaming text according to punctuation marks and line breaks to obtain the current frame text;

[0035] Speech synthesis module: used to perform speech synthesis on the current frame text to obtain audio data corresponding to the current frame text;

[0036] Audio decoding and playing module: used for performing audio decoding on the audio data to obtain decoded audio data, calculating the playing time of the decoded audio data, and playing the decoded audio data;

[0037] Loop module: used to set the time before playing the decoded audio data according to the playing time, and repeat the streaming text segmentation, speech synthesis, audio decoding and audio playing processing until the streaming text speech synthesis and broadcasting are completed.

[0038] Optionally, the text processing module, the audio decoding and playing module and the loop module are all deployed on the client.

[0039] Optionally, the speech synthesis module includes a client speech synthesis sub-module and a cloud speech synthesis sub-module, the client speech synthesis sub-module is deployed on the client, the client speech synthesis sub-module is used to send the current frame text to the cloud speech synthesis sub-module, and receive the audio data corresponding to the current frame text returned by the cloud speech synthesis sub-module; the cloud speech synthesis sub-module is deployed on the cloud, and the cloud speech synthesis sub-module is used to perform speech synthesis on the current frame text to generate audio data corresponding to the current frame text.

[0040] In the present invention, a streaming text that needs to be voice-synthesized and broadcast is received, and the streaming text is segmented according to punctuation marks and line breaks to obtain a current frame text; voice synthesis is performed on the current frame text to obtain audio data corresponding to the current frame text; audio decoding is performed on the audio data to obtain decoded audio data, the playback time of the decoded audio data is calculated, and the decoded audio data is played; according to the playback time, a time is set before the decoded audio data is played, and the streaming text segmentation, voice synthesis, audio decoding and audio playback processing are repeated until the voice synthesis and broadcasting of the streaming text are completed. The present invention deploys a text segmentation strategy on a client to improve the applicability of the client, supports streaming input text in the upstream and streaming return audio data in the downstream; segments the streaming text according to punctuation marks and line breaks, adds a line break identifier, ensures that the streaming text can be segmented even without punctuation marks, and sets a time before playing the audio data corresponding to the current frame text to start segmenting the next frame text, so as to quickly obtain the segmented text that requires speech synthesis, greatly reduce the playback waiting time, and maximize the integrity of the text while ensuring smooth playback; shares a player for playback, reduces the pause between sentences, ensures smoother voice playback, reduces unnecessary system overhead, and facilitates the management of playback status; is suitable for streaming text scenarios based on large model returns, and solves the problems of large voice broadcast delay, severe pause, and poor user experience in large model scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a preferred embodiment of the processing method based on streaming text-to-speech synthesis and broadcasting of the present invention;

[0042] Figure 2 is another flow chart of a preferred embodiment of the processing method based on streaming text-to-speech synthesis and broadcasting of the present invention;

[0043] Figure 3 It is a structural diagram of a preferred embodiment of the processing system based on streaming text-to-speech synthesis and broadcasting of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0045] Existing smart TVs have the ability to use large-model voice. After a question is input, in order to improve the response speed, the background will not return the entire text content for the question at one time, but will stream the text, and then broadcast the streamed text. Existing speech synthesis and broadcasting solutions are generally implemented through the cloud. Downlink supports streaming return of audio data, but uplink does not support streaming input text. After receiving all the texts that need speech synthesis, the texts are sent to the background together. The background streams the audio data after speech synthesis, and then the audio data is played using a player. Obviously, in a large-model scenario, if you wait for all the texts to be returned before initiating the speech synthesis process, you will have to wait for a long time, and the experience will be poor. Therefore, the existing speech synthesis and broadcasting solutions cannot complete the speech synthesis and broadcasting of streaming text well, and there are problems such as large delay, severe pause, and poor user experience.

[0046] In addition, existing speech synthesis and broadcasting solutions also include: simply segmenting the text according to periods, question marks, and exclamation marks, and then synthesizing and broadcasting each sentence of text separately. After the speech synthesis and broadcasting of a sentence is completed, the speech synthesis and broadcasting of the next sentence of text is initiated. However, due to the overly simple segmentation strategy, the user still has a large delay in hearing the played sound, and a new request is initiated for each sentence, and the waiting time between sentences is long, which is not a good user experience.

[0047] In order to solve the above technical problems, the present invention provides a processing method based on streaming text speech synthesis and broadcasting, which receives streaming text that needs speech synthesis and broadcasting; segments the streaming text according to punctuation marks and line breaks to obtain current frame text; performs speech synthesis on the current frame text to obtain audio data corresponding to the current frame text; performs audio decoding on the audio data to obtain decoded audio data, and calculates the playback time of the decoded audio data; plays the decoded audio data; according to the playback time, sets a time before the decoded audio data is played, and repeats streaming text segmentation, speech synthesis, audio decoding and audio playback processing until the streaming text speech synthesis and broadcasting are completed. The present invention deploys a text segmentation strategy on a client to improve the applicability of the client, supports streaming input text in the upstream and streaming return audio data in the downstream; segments the streaming text according to punctuation marks and line breaks, adds a line break identifier, ensures that the streaming text can be segmented even without punctuation marks, and sets a time before playing the audio data corresponding to the current frame text to start segmenting the next frame text, so as to quickly obtain the segmented text that requires speech synthesis, greatly reduce the playback waiting time, and maximize the integrity of the text while ensuring smooth playback; shares a player for playback, reduces the pause between sentences, ensures smoother voice playback, reduces unnecessary system overhead, and facilitates the management of playback status; is suitable for streaming text scenarios based on large model returns, and solves the problems of large voice broadcast delay, severe pause, and poor user experience in large model scenarios.

[0048] The application content is further explained below through the description of embodiments in conjunction with the accompanying drawings.

[0049] The preferred embodiment of the processing method based on streaming text speech synthesis and broadcasting of the present invention is as follows Figure 1 and Figure 2 As shown, specifically including:

[0050] S1. Receive a streaming text that needs to be voice synthesized and broadcasted, and segment the streaming text according to punctuation marks and line breaks to obtain a current frame text.

[0051] In an implementation of this embodiment, the receiving of the streaming text requiring speech synthesis and broadcasting is specifically:

[0052] Receive streaming text output by the large model that requires speech synthesis and broadcasting through the client;

[0053] The streaming text includes text in Markdown format.

[0054] Specifically, existing smart TVs have the ability of large-model voice. After a question is input, in order to improve the response speed, the background will not return the entire text content for the question at one time, but will return the text in a streaming manner, that is, it will return the text in a streaming Markdown format. The present invention receives the Markdown-formatted text that needs to be voice synthesized and broadcasted output by the large model through the client, and is suitable for the streaming text voice synthesis and broadcasting scenarios returned by the large model. Markdown is a lightweight markup language that allows people to write documents in an easy-to-read and easy-to-write plain text format, and then convert it into a valid XHTML (Extensible HyperText Markup Language) or HTML (HyperText Markup Language) document.

[0055] In an implementation of this embodiment, segmenting the streaming text according to punctuation marks and line breaks to obtain the current frame text specifically includes:

[0056] On the client, using punctuation marks and line breaks as identifiers, querying the received streaming text from back to front;

[0057] If the identifier is found, the identifier and the text before the identifier are segmented to obtain the current frame text;

[0058] The current frame text is sent to the cloud.

[0059] Specifically, the present invention deploys the text segmentation strategy on the client to improve the applicability of the client, and realizes the support of streaming input text in the upstream and streaming return audio data in the downstream. The current mainstream cloud-based speech synthesis solution only supports downstream streaming. Putting the text segmentation strategy on the client converts the solution that only supports downstream streaming speech synthesis into a speech synthesis solution that supports streaming in both upstream and downstream, which is suitable for the scenario of streaming text speech synthesis and broadcasting returned by large models. The text segmentation strategy of the present invention is: for the first text segmentation, search from back to front according to Chinese and English punctuation marks and line breaks (".", ","", "?"", "!"", ";", ":", "."", ","", "?"", "!"", "\n"), if the above symbols are found, take the symbols and the previous text, and determine it as the text for the first segmentation, that is, the first frame text; if not found, wait for the next large model to return the streaming text, and then check; for subsequent text segmentation, first search from back to front according to ".", "?", "!"", "?"", "!"", "\n", if the above symbols are found, take the symbols and the previous text, and determine it as the text that needs to be segmented, if not found, then search according to ".", "," "?"", "!"", ";", ":", ".", "," "?"", "\n". The present invention creatively adds line breaks as segmentation marks, solves the situation that there is no punctuation in the Markdown format file and also needs to be segmented, gets the text that needs speech synthesis in the first frame as soon as possible, and initiates speech synthesis, greatly reducing the playback waiting time. At the same time, except for the first segmented text, try to ensure the maximum length of the subsequent segmented text. This can ensure the integrity of the Markdown formatted text as much as possible without causing delays, making it easier for the cloud-based speech synthesis engine to process the Markdown formatted text.

[0060] S2. Perform speech synthesis on the current frame text to obtain audio data corresponding to the current frame text.

[0061] In an implementation of this embodiment, performing speech synthesis on the current frame text to obtain audio data corresponding to the current frame text specifically includes:

[0062] In the cloud, using a TTS method to perform speech synthesis on the current frame text to obtain audio data corresponding to the current frame text;

[0063] Returning the audio data to the client;

[0064] Wherein, the audio data includes audio data in mp3 format.

[0065] Specifically, after a TTS speech synthesis request is initiated to the cloud-based speech synthesis engine, the cloud-based speech synthesis engine is responsible for receiving the text that requires speech synthesis and returning streaming audio data, namely, a TTS audio stream.

[0066] S3. Perform audio decoding on the audio data to obtain decoded audio data, calculate the playing time of the decoded audio data, and play the decoded audio data.

[0067] In an implementation of this embodiment, the audio decoding of the audio data to obtain the decoded audio data and calculating the playback time of the decoded audio data specifically includes:

[0068] On the client, audio decoding is performed on the audio data (audio in mp3 format) to obtain decoded audio data (audio in pcm format), wherein the decoded audio data includes audio data in pcm format;

[0069] Calculate the playback duration of the decoded audio data:

[0070] T = L × 8 / (F × A × B);

[0071] Among them, T represents the playback time, in seconds; L represents the size of the decoded audio data, that is, the file size, in bytes; F represents the sampling rate, in Hz; A represents the number of sampling bits, in bits; and B represents the number of channels.

[0072] In an implementation of this embodiment, the playing of the decoded audio data is specifically:

[0073] On the client, playing the decoded audio data, i.e., the audio in pcm format, through a player;

[0074] Among them, the different decoded audio data corresponding to texts with different frame numbers are played by a common player.

[0075] Specifically, the present invention shares a player, reduces the pauses between sentences, and makes the synthesized speech playback smoother. That is, only one player is created for a streaming text-to-speech synthesis. While reducing the problem of unsmooth playback, it can also reduce unnecessary system overhead and facilitate the management of playback status.

[0076] S4. According to the playing time, a time is set before the decoded audio data is played, and streaming text segmentation, speech synthesis, audio decoding and audio playing processing are repeated until the streaming text speech synthesis and broadcasting are completed.

[0077] In an implementation of this embodiment, according to the playback duration, a time is set before the decoded audio data is played, and streaming text segmentation, speech synthesis, audio decoding and audio playback processing are repeated until the streaming text speech synthesis and broadcasting are completed, specifically including:

[0078] According to the playing time, a time is set before the decoded audio data is played (set to 2s in this embodiment), and the streaming text except the current frame text is segmented according to punctuation marks and line breaks to obtain the next frame text;

[0079] Performing speech synthesis, audio decoding and audio playback processing on the next frame of text;

[0080] The streaming text segmentation, speech synthesis, audio decoding and audio playing processes are repeated until the entire streaming text speech synthesis and broadcasting are completed.

[0081] Specifically, Figure 2 As shown, in some scenarios where speech synthesis needs to be initiated but all texts that need to synthesize audio cannot be obtained, such as the streaming text scenario of the large model streaming return, after obtaining the text, the speech synthesis process is initiated, and the text is segmented after the punctuation or line break is detected for the first time, and the segmented text is sent to the cloud. The cloud converts the text into streaming audio data and returns it, and then decodes the audio data into pcm original data, and plays the audio according to the decoded audio data. And by the length of the decoded audio file, the duration of the audio file playback is calculated. When the audio is played before 2s, the query starts from the back of the text. If punctuation or line breaks can be found, the punctuation or line breaks and all the texts before them are taken out and sent to the cloud again for speech synthesis, and then decoded and played. Repeat this process until all the texts that need to be played are played. The present invention starts the segmentation of the next frame of text 2s before the audio data corresponding to the current frame of text is played, so as to quickly obtain the segmented text that needs speech synthesis, greatly reduce the waiting time for playback, and ensure the smooth playback while also maximizing the integrity of the text.

[0082] The creativity of the present invention lies in segmenting the text according to punctuation marks and line breaks, especially introducing line breaks during segmentation, so as to minimize the sense of delay in broadcasting caused by the delay under the large model streaming text. At the same time, in addition to the text segmented for the first time, the maximum length of the subsequent segmented text is guaranteed as much as possible, so that the integrity of the Markdown formatted text can be guaranteed as much as possible without generating delays, which is convenient for the cloud-based speech synthesis engine to process the Markdown formatted text. By adding a segmentation scheme to the client, the speech synthesis scheme that does not support upstream streaming can be transformed into an upstream and downstream streaming speech synthesis method. In addition, a streaming text speech synthesis will only create one player, which can reduce the problem of unsmooth playback while also reducing unnecessary system overhead and facilitating the management of playback status. The present invention implements a speech synthesis and playback scheme for streaming uploaded text, streaming returned audio data, and played, to perfectly adapt to the speech broadcast in the streaming text scenario based on the large model return, thereby solving the problem of large voice broadcast delay and serious pause in the large model scenario.

[0083] In addition, based on the above-mentioned processing method based on streaming text speech synthesis and broadcasting, the present invention also provides a processing system based on streaming text speech synthesis and broadcasting, wherein a preferred embodiment of the processing system based on streaming text speech synthesis and broadcasting is as follows: Figure 3 As shown, specifically including:

[0084] Text processing module 01: used to receive streaming text that needs to be synthesized and broadcast, and segment the streaming text according to punctuation marks and line breaks to obtain the current frame text;

[0085] Speech synthesis module 02: used to perform speech synthesis on the current frame text to obtain audio data corresponding to the current frame text;

[0086] Audio decoding and playing module 03: used for performing audio decoding on the audio data to obtain decoded audio data, calculating the playing time of the decoded audio data, and playing the decoded audio data;

[0087] Loop module 04: used to set a time before playing the decoded audio data according to the playing time, and repeat the streaming text segmentation, speech synthesis, audio decoding and audio playing processing until the streaming text speech synthesis and broadcasting are completed.

[0088] In an implementation of this embodiment, the text processing module, the audio decoding and playing module, and the loop module are all deployed on the client.

[0089] In one implementation of this embodiment, the speech synthesis module includes a client speech synthesis sub-module and a cloud speech synthesis sub-module. The client speech synthesis sub-module is deployed on the client, and the client speech synthesis sub-module is used to send the current frame text to the cloud speech synthesis sub-module, and receive the audio data corresponding to the current frame text returned by the cloud speech synthesis sub-module; the cloud speech synthesis sub-module is deployed on the cloud, and the cloud speech synthesis sub-module is used to perform speech synthesis on the current frame text to generate audio data corresponding to the current frame text.

[0090] In one implementation of this embodiment, the text processing module includes a text receiving submodule and a text segmentation submodule. The text receiving submodule is used to receive streaming text that needs to be voice synthesized and broadcasted; the text segmentation submodule is used to segment the streaming text according to punctuation marks and line breaks to obtain the current frame text.

[0091] In one implementation of this embodiment, the audio decoding and playback module includes an audio decoding submodule and an audio playback submodule, the audio decoding submodule is used to perform audio decoding on the audio data, obtain decoded audio data, and calculate the playback time of the decoded audio data; the audio playback submodule is used to play the decoded audio data.

[0092] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal 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, article or terminal. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or terminal including the element.

[0093] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium that can be read by a computer, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0094] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A processing method based on streaming text-to-speech synthesis and broadcasting, characterized in that: The processing method based on streaming text-to-speech synthesis and broadcasting includes: Receiving a stream text that needs to be synthesized and broadcasted, and segmenting the stream text according to punctuation marks and line breaks to obtain a current frame text; Performing speech synthesis on the current frame text to obtain audio data corresponding to the current frame text; Performing audio decoding on the audio data to obtain decoded audio data, calculating a playback duration of the decoded audio data, and playing the decoded audio data; According to the playing time, a time is set before the decoded audio data is played, and the streaming text segmentation, speech synthesis, audio decoding and audio playing processing are repeated until the streaming text speech synthesis and broadcasting are completed.

2. The processing method based on streaming text-to-speech synthesis and broadcasting according to claim 1, characterized in that: The receiving of the streaming text that needs to be voice synthesized and broadcasted is specifically: Receive the streaming text output by the large model through the client that needs to be synthesized and broadcasted; The streaming text includes text in Markdown format.

3. The processing method based on streaming text-to-speech synthesis and broadcasting according to claim 2, characterized in that: The step of segmenting the stream text according to punctuation marks and line breaks to obtain the current frame text specifically includes: On the client, using punctuation marks and line breaks as identifiers, querying the received streaming text from back to front; If the identifier is found, the identifier and the text before the identifier are segmented to obtain the current frame text; The current frame text is sent to the cloud.

4. The processing method based on streaming text-to-speech synthesis and broadcasting according to claim 3 is characterized in that: The performing speech synthesis on the current frame text to obtain audio data corresponding to the current frame text specifically includes: In the cloud, using a TTS method to perform speech synthesis on the current frame text to obtain audio data corresponding to the current frame text; Returning the audio data to the client; Wherein, the audio data includes audio data in mp3 format.

5. The processing method based on streaming text-to-speech synthesis and broadcasting according to claim 2, characterized in that: The step of performing audio decoding on the audio data to obtain decoded audio data and calculating the playback duration of the decoded audio data specifically includes: On the client, performing audio decoding on the audio data to obtain decoded audio data, wherein the decoded audio data includes audio data in a pcm format; Calculate the playback duration of the decoded audio data: T = L × 8 / (F × A × B); Among them, T represents the playback time, L represents the size of the decoded audio data, F represents the sampling rate, A represents the number of sampling bits, and B represents the number of channels.

6. The processing method based on streaming text-to-speech synthesis and broadcasting according to claim 5, characterized in that: The playing of the decoded audio data is specifically: On the client, playing the decoded audio data through a player; Among them, the different decoded audio data corresponding to texts with different frame numbers are played by a common player.

7. The processing method based on streaming text-to-speech synthesis and broadcasting according to claim 1, characterized in that: The method of setting a time before playing the decoded audio data according to the playing time, repeating the streaming text segmentation, speech synthesis, audio decoding and audio playing processing until the streaming text speech synthesis and broadcasting are completed, specifically includes: According to the playing time, a time is set before the decoded audio data is played, and the streaming text except the current frame text is segmented according to punctuation marks and line breaks to obtain the next frame text; Performing speech synthesis, audio decoding and audio playback processing on the next frame of text; The streaming text segmentation, speech synthesis, audio decoding and audio playing processes are repeated until the entire streaming text speech synthesis and broadcasting are completed.

8. A processing system based on streaming text-to-speech synthesis and broadcasting, characterized in that: The processing system based on streaming text speech synthesis and broadcasting includes: Text processing module: used for receiving streaming text that needs to be synthesized and broadcast, segmenting the streaming text according to punctuation marks and line breaks to obtain the current frame text; Speech synthesis module: used to perform speech synthesis on the current frame text to obtain audio data corresponding to the current frame text; Audio decoding and playing module: used for performing audio decoding on the audio data to obtain decoded audio data, calculating the playing time of the decoded audio data, and playing the decoded audio data; Loop module: used to set the time before playing the decoded audio data according to the playing time, and repeat the streaming text segmentation, speech synthesis, audio decoding and audio playing processing until the streaming text speech synthesis and broadcasting are completed.

9. The processing system based on streaming text-to-speech synthesis and broadcasting according to claim 8, characterized in that: The text processing module, the audio decoding and playing module and the loop module are all deployed on the client.

10. The processing system based on streaming text-to-speech synthesis and broadcasting according to claim 8, characterized in that: The speech synthesis module includes a client speech synthesis submodule and a cloud speech synthesis submodule. The client speech synthesis submodule is deployed on the client, and is used to send the current frame text to the cloud speech synthesis submodule, and receive the audio data corresponding to the current frame text returned by the cloud speech synthesis submodule; the cloud speech synthesis submodule is deployed on the cloud, and is used to perform speech synthesis on the current frame text to generate audio data corresponding to the current frame text.

Images