Audio detection method and audio detection device

Abstract

An audio detection method executed by a processor reading at least command stored in a memory is disclosed herein. The audio detection method includes following steps: performing a framing process on an audio to generate audio frames; segmenting the audio into a plurality of audio segments based on signal intensities of the audio frames; calculating a plurality of actual pitches of the audio frames in a target audio segment of the plurality of audio segments; calculating a plurality of predicted pitches based on the plurality of actual pitches; calculating a plurality of pitch differences based on the plurality of actual pitches and the plurality of predicted pitches; calculating a threshold exceedance count based on the plurality of pitch differences and a pitch threshold; determining the target audio segment as a multiple sound source audio segment if the threshold exceedance count is larger than a portion of a total number of pitches of the plurality of actual pitches; and determining the target audio segment as a single sound source audio segment if the threshold exceedance count is zero.

Description

Technical Field

[0001] This case relates to audio detection methods and devices, and in particular to audio detection methods and devices for detecting audio to obtain the number of sound sources. Prior Technology

[0002] In the event of an accident, the elderly and school-aged children are more in need of assistance. Using electronic devices to detect accidents and notifying emergency contacts of the results will help them understand the situation immediately and take appropriate action.

[0003] However, existing detection methods typically utilize image tracking technology, which requires the installation of additional cameras. Furthermore, in cases of falls, fainting, or other unexpected situations, it is difficult to use image tracking technology to monitor the actual accidental situations of the elderly and school-aged children. Summary of the Invention

[0004] In view of the shortcomings of prior art, one of the objectives of this application is (but not limited to) to provide an audio detection method and an audio detection device to improve the shortcomings of prior art.

[0005] In some embodiments, this invention provides an audio detection method executed by a processor reading at least one instruction stored in memory. The audio detection method includes the following steps: framing the audio signal to generate a plurality of sound frames; dividing the audio signal into a plurality of segments based on a plurality of signal intensities of the plurality of sound frames; calculating a plurality of actual pitches of the plurality of sound frames of a target segment; calculating a plurality of predicted pitches based on the plurality of actual pitches of the target segment; calculating a plurality of pitch differences based on the plurality of actual pitches and the plurality of predicted pitches of the target segment; calculating an over-threshold count based on the plurality of pitch differences of the target segment and a pitch threshold; if the over-threshold count is greater than a portion of the total number of pitches of the plurality of actual pitches of the target segment, the target segment is a plurality of sound source segments; and if the over-threshold count is zero, the target segment is a single sound source segment.

[0006] In some embodiments, the audio detection device includes a memory and a processor. The memory stores at least one instruction. The processor reads the at least one instruction to perform the following steps: framing the audio to generate a plurality of sound frames; dividing the audio into a plurality of segments based on a plurality of signal intensities of the plurality of sound frames; calculating a plurality of actual pitches of the plurality of sound frames of the target segment; calculating a plurality of predicted pitches based on the plurality of actual pitches of the target segment; calculating a plurality of pitch differences based on the plurality of actual pitches and the plurality of predicted pitches of the target segment; calculating an over-threshold count based on the plurality of pitch differences of the target segment and a pitch threshold; if the over-threshold count is greater than a portion of the total number of pitches of the plurality of actual pitches of the target segment, the target segment is a plurality of sound source segments; and if the over-threshold count is zero, the target segment is a single sound source segment.

[0007] The technical means embodied in the embodiments of this case can improve at least one of the shortcomings of the prior art. The audio detection method and audio detection device of this case can be applied to various electronic systems. By receiving audio through the electronic system and calculating the audio, the number of users can be determined. This not only avoids the problem of blind spots in image tracking technology, which makes it impossible to effectively monitor users, but also has the advantages of low computational load and simple implementation.

[0008] Regarding the features, implementation, and effects of this case, the preferred embodiments are described in detail below with reference to the drawings. Simple Explanation of the Diagram

[0009] Figure 1 is a schematic diagram of an audio detection device according to some embodiments of this case; Figure 2 is a flowchart of an audio detection method according to some embodiments of this case; Figure 3 is a schematic diagram of an audio message and a sound frame according to some embodiments of this case; Figure 4 is a schematic diagram of an audio signal and pitch according to some embodiments of this case; Figure 5 is a schematic diagram of an audio signal and pitch according to some embodiments of this case; Figure 6 is a flowchart of an audio detection method according to some embodiments of this case; Figure 7 is a schematic diagram of an audio signal and pitch according to some embodiments of this case; Figure 8 is a schematic diagram illustrating an audio signal and pitch according to some embodiments of this case; and Figure 9 is a schematic diagram of an audio signal and pitch according to some embodiments of this case. Implementation

[0010] All terms used herein have their common meanings. The definitions of the terms mentioned above in commonly used dictionaries, and any examples of the use of any term discussed herein, are merely illustrative and should not limit the scope or meaning of this application. Similarly, this application is not limited to the various embodiments shown in this specification.

[0011] As used herein, "coupling" or "connection" can refer to two or more components making direct physical or electrical contact with each other, or indirectly making direct physical or electrical contact with each other, or to two or more components operating or moving together. As used herein, the term "circuit" can refer to a device in which at least one transistor and / or at least one active or passive component are connected in a certain manner to process signals.

[0012] As used herein, the term "and / or" includes any combination of one or more of the listed related items. In this document, the terms first, second, third, etc., are used to describe and distinguish individual elements. Therefore, a first element in this document may also be referred to as a second element without departing from the intent of this case. For ease of understanding, similar elements in the various diagrams will be assigned the same reference numerals.

[0013] To avoid the problem of blind spots in video tracking technology that prevent effective user monitoring, this case proposes an audio detection method and an audio detection device, detailed below.

[0014] Figure 1 is a schematic diagram of an audio detection device 100 according to some embodiments of the present invention. As shown, the audio detection device 100 includes a microphone 110, an audio input interface 120, a processor 130, and a memory 140. The microphone 110 is used to receive sound and convert the sound into digital audio. Subsequently, the microphone 110 inputs the digital audio to the processor 130 through the audio input interface 120. The memory 140 is used to store at least one instruction. The processor 130 is used to read at least one instruction to execute the audio detection process. To make the operation of the audio detection device 100 easier to understand, please also refer to Figure 2, which is a flowchart of an audio detection method 200 according to some embodiments of the present invention.

[0015] Please refer to step 210 in Figure 2, where the audio is framing to generate a plurality of sound frames. For example, referring to Figures 1 and 3, the processor 130 of the audio detection device 100 of this invention can framing the audio to generate a plurality of sound frames fn~fn+2. In detail, the processor 130 of the audio detection device 100 of this invention can divide the audio into multiple overlapping sound frames fn~fn+2, and then perform further acoustic analysis on a unit of sound frames fn~fn+2.

[0016] The sampling rate of the processor 130 of the audio detection device 100 in this invention varies depending on the audio input interface 120, and can be 16000Hz, 24000Hz, 44100Hz, or 48000Hz. In some embodiments, the frame size of the audio frame can be 20 milliseconds (ms). If the sampling rate is 48000Hz, then 960 samples are sampled. The overlap rate between audio frames can be 25%, and the overlap time between audio frames is 5 milliseconds (ms), resulting in 240 overlapping samples between audio frames. Each audio frame is separated by a hop size, which can be the frame size minus the overlap time. For example, 20 milliseconds minus 5 milliseconds equals 15 milliseconds, and the hop size contains 720 samples. However, this invention is not limited to the values ​​given in this embodiment, and other suitable values ​​can also be used, depending on the actual needs.

[0017] Please refer to step 220 in Figure 2, where the audio signal is divided into multiple segments based on the multiple signal intensities of multiple sound frames. For example, referring to Figures 1 and 3, the processor 130 of the audio detection device 100 in this case can calculate the signal intensity of sound frames fn~fn+2, and the formula for the signal intensity is as follows:

[0018] …Formula 1

[0019] As shown in Formula 1, It's the signal strength. is the sample value of the sound frame, and N is the frame size of the sound frame.

[0020] Furthermore, referring to Figures 1 and 4, the processor 130 of the audio detection device 100 of this invention can obtain a plurality of peaks p1~p10 and a plurality of troughs t1~t10 based on a plurality of signal strengths. For example, to obtain a peak, the signal strength of the sound frame must first be calculated, and the signal strength of the sound frame must be greater than the signal strength of the sound frames on both sides of it. In this way, the sound frame with the largest relative signal strength can be found as a candidate peak. Moreover, this invention sets thresholds to filter candidate peaks, such as setting a maximum height, a minimum height, a minimum distance between peaks, etc. In detail, this invention can set a minimum height of 1dB, a maximum height of 25dB, and a minimum distance between peaks of 5 hop sizes. If the hop size is 15 milliseconds, then the minimum distance between peaks is 75 milliseconds. Furthermore, the method for obtaining the troughs is similar to that for obtaining the peaks, the only difference being that the signal strength of the sound frame must be changed to a negative value (multiplied by -1) for the relevant calculations. The other calculation methods are the same and will not be elaborated here. The processor 130 of the audio detection device 100 in this case can obtain the peak array as follows according to the above method:

[0021] …Formula 2

[0022] As shown in Formula 2, The parentheses indicate the peak, and the numbers within them represent the frame number. For example, 12 in parentheses indicates the 12th frame, meaning the 12th frame is a peak. Similarly, 30 in parentheses indicates the 30th frame, meaning the 30th frame is a peak, and so on.

[0023] The processor 130 of the audio detection device 100 in this case can obtain the valley array as described above:

[0024] …Formula 3

[0025] As shown in Formula 3, The parentheses indicate the trough, and the numbers within them represent the note frame number. For example, 1 within the parentheses indicates the first note frame, meaning the first note frame is a trough. Similarly, 23 within the parentheses indicates the 23rd note frame, meaning the 23rd note frame is a trough, and so on.

[0026] Furthermore, the processor 130 of the audio detection device 100 in this case can divide the audio signal into multiple segments s1 to s9 based on a plurality of troughs t1 to t10. For example, the processor 130 of the audio detection device 100 in this case can divide the audio signal into segment s1 based on troughs t1 and t2, and the processor 130 can divide the audio signal into segment s2 based on troughs t2 and t3, and so on. In detail, assuming that trough t1 is the first trough in the trough array of Formula 3, according to Formula 3, trough t1 is the first sound frame. In addition, assuming that trough t2 is the second trough in the trough array of Formula 3, according to Formula 3, trough t2 is the 23rd sound frame. The calculation formula for segment s1 divided by troughs t1 and t2 is as follows:

[0027] …Formula 4

[0028] …Formula 5

[0029] As shown in formulas 4 and 5, For the time point t1 of the trough, For the time point t2 of the trough, This is the first trough in the trough array of Formula 3. This is the second trough in the trough array of Formula 3. This is the jump distance. Assume... For the first sound frame, the jump distance is 15 milliseconds. Substituting into Formula 4, we can obtain the time point of the trough t1 as 15 milliseconds. Assuming... For the 23rd sound frame, with a jump interval of 15 milliseconds, substituting into Formula 5, the time point of the trough t2 is found to be 345 milliseconds. Therefore, the sound segment s1, divided by troughs t1 and t2, falls between 15 milliseconds and 345 milliseconds. The processor 130 of the audio detection device 100 of this invention can obtain the time intervals of all sound segments s1 to s9 in the manner described above. However, this invention is not limited to the values ​​given in this embodiment; other suitable values ​​may also be used, depending on actual needs.

[0030] Please refer to step 230 in Figure 2 to calculate the multiple actual pitches of the multiple sound frames of the target sound segment for the multiple sound segments. For example, referring to Figures 1 and 4, the processor 130 of the audio detection device 100 of this invention can calculate the multiple actual pitches of the multiple sound frames of the target sound segment s1 for the multiple sound segments s1 to s9. In detail, the processor 130 of the audio detection device 100 of this invention can use the YIN algorithm to calculate the actual pitches of the sound frames in the sound segment s1. However, this invention is not limited to the YIN algorithm described in this embodiment, and other suitable algorithms may also be used, depending on the actual needs.

[0031] Please refer to step 240 in Figure 2, where a plurality of predicted pitches are calculated based on a plurality of actual pitches of the target audio segment. For example, referring to Figures 1 and 4, the processor 130 of the audio detection device 100 of this invention can calculate the predicted pitch based on the actual pitch of the sound frame in the target audio segment s1. In detail, the processor 130 of the audio detection device 100 of this invention can use Kalman filter tracking to calculate the predicted pitch, and the formula for the predicted pitch is as follows:

[0032] …Formula 6

[0033] As shown in Formula 6, The actual pitch at time t, The predicted pitch for time t, The predicted pitch at time t-1, This refers to the Kalman gain. In this invention, the predicted pitch can be calculated from the actual pitch of the tone frame in the target pitch segment s1 using Formula 6. Please refer to Figure 5 for the actual pitch and the calculated predicted pitch; the solid line represents the actual pitch, and the dashed line represents the predicted pitch calculated using Formula 6. However, this invention is not limited to the Kalman filter tracking described in this embodiment; other suitable algorithms can also be used, depending on the actual requirements.

[0034] Please refer to step 250 in Figure 2, where a plurality of pitch differences are calculated based on a plurality of actual pitches and a plurality of predicted pitches of the target audio segment. For example, referring to Figures 1 and 5, the processor 130 of the audio detection device 100 of this invention can calculate the pitch difference based on the actual pitch (solid line) and predicted pitch (dashed line) of the target audio segment. The formula for the pitch difference is as follows:

[0035] …Formula 7

[0036] As shown in Formula 7, For pitch difference, This is the actual pitch. To predict pitch, refer to Figure 5. If the actual pitch is close to the predicted pitch, it indicates a single sound source (e.g., only one person speaking). In this case, the actual pitch is close to the predicted pitch, and the pitch difference is small. Conversely, if the actual pitch is far from the predicted pitch, it indicates that the prediction is inaccurate because there is more than one sound source (e.g., multiple people speaking). In this case, the actual pitch deviates from the predicted pitch, and the pitch difference is larger.

[0037] Please refer to step 260 in Figure 2, where the threshold exceedance count is calculated based on the multiple pitch differences and pitch thresholds of the target audio segment. For example, referring to Figure 1, the processor 130 of the audio detection device 100 in this case can calculate the threshold exceedance count based on the pitch differences and pitch thresholds of the target audio segment. The formula for the threshold exceedance count is as follows:

[0038] …Formula 8

[0039] As shown in Formula 8, For counting exceeding the threshold, For pitch difference, For pitch threshold, 1(⋅) in Formula 8 is an indicator function, which is 1 when the condition in parentheses is true, and 0 otherwise.

[0040] Please refer to step 270 in Figure 2. If the threshold count is greater than a portion of the total number of actual pitches in the target audio segment, the target audio segment is a plurality of sound source segments. For example, referring to Figures 1 and 4, the processor 130 of the audio detection device 100 in this case can use the YIN algorithm to calculate the actual pitch of the sound frames in audio segment s1. Assuming the YIN algorithm calculates that sound frame 1 has an actual pitch, then sound frame 1 is set as a sound frame with pitch. Assuming the YIN algorithm calculates that sound frame 2 does not have an actual pitch, then sound frame 2 is set as a sound frame without pitch. In this case, pitch detection can be performed on all sound frames in audio segment s1 in the above manner, and the number of sound frames with pitch is the total number of pitches. In addition, if the threshold count S in Formula 8 is greater than half of the total number of pitches, then the target audio segment s1 is determined to be a plurality of sound source segments, which means that multiple people are speaking in the target audio segment s1.

[0041] Please refer to step 280 in Figure 2. If the threshold count is zero, the target audio segment is a single-source audio segment. For example, please refer to Figures 1 and 4. If the threshold count S in Formula 8 is 0, then the target audio segment s1 is determined to be a single-source audio segment, meaning that only one person is speaking in the target audio segment s1. As described above, the audio detection device 100 and audio detection method 200 of this case can receive audio through the microphone 110 and calculate the audio to determine the number of users on site. In this way, the audio detection device 100 and audio detection method 200 of this case can not only avoid the problem of blind spots in image tracking technology that prevent effective monitoring of users, but also have the advantages of low computational load and simple implementation.

[0042] Figure 6 is a flowchart of an audio detection method 600 according to some embodiments of this invention. This invention can use the audio detection method 600 to determine all the audio segments s1~s9 in Figure 4, thereby identifying which type of audio segment s1~s9 belongs to, and thus understanding how many speakers are in each audio segment s1~s9.

[0043] Please refer to step 610 in Figure 6. The audio detection method 600 in this case can determine whether a sound segment has no peaks. Assuming that the sound segment s1 does indeed have no peaks, then in step 611, the sound segment s1 is determined to be a sound source-less sound segment; in other words, there is no speaker in the sound segment s1. Alternatively, assuming that the sound segment s1 has peaks, then step 620 is executed.

[0044] Please refer to step 620 in Figure 6. The audio detection method 600 in this case can determine whether the total number of pitches is zero. Assuming that the total number of pitches is zero, then in step 621, it is determined that the pitch segment s1 is a breathy pitch segment; in other words, the pitch segment s1 contains only breathy sounds. Furthermore, assuming that the total number of pitches in the pitch segment s1 is not zero, then step 630 is executed.

[0045] Please refer to step 630 in Figure 6. The audio detection method 600 of this invention can determine whether the signal strength of the boundary frame is greater than the intensity threshold. Assuming that the range of the sound segment s1 is between the first frame and the 23rd frame, then the first frame and the 23rd frame are both boundary frames of the sound segment s1. The audio detection method 600 of this invention can set the intensity threshold to -3dB. If the signal strength of the first frame is greater than -3dB or the signal strength of the 23rd frame is greater than -3dB, then in step 631, it is determined that the sound segment s1 is a plurality of sound source segments. In other words, there are multiple speakers in the sound segment s1. Furthermore, assuming that the signal strength of neither the first frame nor the 23rd frame is greater than -3dB, then step 640 is executed.

[0046] Please refer to step 640 in Figure 6. The audio detection method 600 of this case can determine whether the threshold count is greater than a portion of the total number of pitches. For example, if the threshold count S of Formula 8 is greater than half of the total number of pitches, then in step 641, it is determined that the sound segment s1 is a plurality of sound source sound segments; in other words, there are multiple speakers in the sound segment s1. Furthermore, if the threshold count S of Formula 8 is not greater than half of the total number of pitches, then step 650 is executed.

[0047] Please refer to step 650 in Figure 6. The audio detection method 600 in this case can determine whether the threshold count is zero. For example, if the threshold count S in Formula 8 is 0, then in step 651, it is determined that the sound segment s1 is a single sound source segment; in other words, there is only one speaker in the sound segment s1. Furthermore, if the threshold count S in Formula 8 is not 0, then step 660 is executed.

[0048] Please refer to step 660 in Figure 6. The audio detection method 600 of this case can determine whether the over-threshold count is between a portion of the total number of pitches and zero. For example, if the over-threshold count S of Formula 8 is between half of the total number of pitches and 0, then in step 661, it is determined that the sound segment s1 is a sound source conversion segment. In other words, the speaker in the sound segment s1 is changing from speaker A to speaker B.

[0049] Figures 7 to 9 are schematic diagrams illustrating audio signals and pitches according to some embodiments of this invention. As shown in Figures 7 to 9, after detecting the audio signal according to the audio detection method 600 of this invention, the number of speakers in different segments can be determined. These segments may contain multiple pitch ranges. For example, in a single-source segment, the actual pitch (solid line) and the predicted pitch (dashed line) are relatively close, indicating that there is currently a single sound source (e.g., only one person speaking). In a multi-source segment, the actual pitch (solid line) and the predicted pitch (dashed line) are far apart, indicating that there is currently more than one sound source, which leads to inaccurate prediction. In this case, there are multiple sound sources (e.g., multiple people speaking). Furthermore, if the state is between a single sound source and multiple sound sources, it indicates that there is currently a sound source transition segment. In other words, the speaker in this segment is transitioning from speaker A to speaker B. Furthermore, if there are no peaks within a segment, it is a sourceless segment; in other words, there is no sound source in the segment (e.g., no one is speaking). Additionally, in the breathy segment, the total number of pitches is zero, indicating that only breathy sounds exist.

[0050] It should be noted that this application is not limited to the embodiments shown in Figures 1 to 9, but is merely used to illustrate one implementation method of this application to facilitate understanding of the technology. The scope of this application shall be determined by the scope of the invention application. Modifications and refinements made by those skilled in the art to the embodiments of this application without departing from the spirit of this application still fall within the scope of the invention application.

[0051] The technical means embodied in the embodiments of this case can improve at least one of the shortcomings of the prior art. The audio detection method and audio detection device of this case can be applied to various electronic systems. By receiving audio through the electronic system and calculating the audio, the number of users can be determined. This not only avoids the problem of blind spots in image tracking technology, which makes it impossible to effectively monitor users, but also has the advantages of low computational load and simple implementation.

[0052] Although the embodiments of this case are described above, these embodiments are not intended to limit this case. Those skilled in the art can make changes to the technical features of this case based on the express or implied content of this case. All such changes may fall within the scope of patent protection sought in this case. In other words, the scope of patent protection in this case shall be determined by the scope of the patent application in this specification.

[0053] 100: Audio Detection Device 110: Radio 120: Audio Input Interface 130: Processor 140: Memory 200: Method 210~280: Steps 600: Method 610~661: Steps fn~ fn+2: Sound frame p1~p10: Peaks s1~s9: Segment t1~t10: troughs

Claims

1. An audio detection method, executed by a processor reading at least one instruction stored in a memory, comprising: performing a frame-segmentation process on an audio signal to generate a plurality of frames; dividing the audio signal into a plurality of segments based on a plurality of signal intensities of the plurality of frames; calculating a plurality of actual pitches of the plurality of frames of a target segment for the plurality of segments; calculating a plurality of predicted pitches based on the plurality of actual pitches of the target segment; calculating a plurality of pitch differences based on the plurality of actual pitches and the plurality of predicted pitches of the target segment; calculating an over-threshold count based on the plurality of pitch differences of the target segment and a pitch threshold; if the over-threshold count is greater than a portion of the total number of pitches of the plurality of actual pitches of the target segment, the target segment is a plurality of sound source segments; and if the over-threshold count is zero, the target segment is a single sound source segment.

2. The audio detection method as described in claim 1 further includes: if the over-threshold count is between half of the total number of pitches in the target audio segment and zero, the target audio segment is a source-converted audio segment.

3. The audio detection method as described in claim 1, wherein dividing the audio into the plurality of segments based on the plurality of signal intensities of the plurality of sound frames comprises: calculating the plurality of signal intensities of the plurality of sound frames; obtaining a plurality of peaks and a plurality of troughs based on the plurality of signal intensities; and dividing the audio into the plurality of segments based on the plurality of troughs.

4. The audio detection method as described in claim 1 further includes: if the target audio segment does not contain any peaks, the target audio segment is a sourceless audio segment; and if the total number of the pitches of the plurality of actual pitches of the target audio segment is zero, the target audio segment is a breathy audio segment.

5. The audio detection method as described in claim 1 further includes: if the intensity of a first signal in a first boundary frame of the target audio segment is greater than an intensity threshold or the intensity of a second signal in a second boundary frame of the target audio segment is greater than the intensity threshold, the target audio segment is the plurality of sound source audio segments.

6. An audio detection device, comprising: a memory for storing at least one instruction; and a processor for reading the at least one instruction from the memory to perform the following steps: performing a frame-segmentation process on an audio signal to generate a plurality of sound frames; dividing the audio signal into a plurality of segments based on a plurality of signal intensities of the plurality of sound frames; calculating a plurality of actual pitches of the plurality of sound frames of a target segment from the plurality of sound segments; calculating a plurality of predicted pitches based on the plurality of actual pitches of the target segment; calculating a plurality of pitch differences based on the plurality of actual pitches and the plurality of predicted pitches of the target segment; and calculating an over-threshold count based on the plurality of pitch differences of the target segment and a pitch threshold. If the threshold count is greater than a portion of the total number of pitches of the plurality of actual pitches of the target pitch segment, the target pitch segment is a plurality of sound source pitch segments; and if the threshold count is zero, the target pitch segment is a single sound source pitch segment.

7. The audio detection device as claimed in claim 6, wherein the processor further reads the at least one instruction from the memory to perform the following steps: if the over-threshold count is between half of the total number of pitches of the target audio segment and zero, the target audio segment is a source-converted audio segment.

8. The audio detection device as claimed in claim 6, wherein the processor further reads the at least one instruction from the memory to perform the following steps: calculating the plurality of signal intensities of the plurality of sound frames; obtaining a plurality of peaks and a plurality of troughs based on the plurality of signal intensities; and dividing the audio signal into the plurality of segments based on the plurality of troughs.

9. The audio detection device as claimed in claim 6, wherein the processor further reads the at least one instruction from the memory to perform the following steps: if the target audio segment does not contain any peaks, the target audio segment is a sourceless audio segment; and if the total number of the pitches of the plurality of actual pitches of the target audio segment is zero, the target audio segment is a breathy audio segment.

10. The audio detection device as claimed in claim 6, wherein the processor further reads the at least one instruction from the memory to perform the following steps: if a first signal strength of a first boundary frame of the target audio segment is greater than a strength threshold or a second signal strength of a second boundary frame of the target audio segment is greater than the strength threshold, the target audio segment is the plurality of sound source segments.

Images