CN100580774C - Method and device for eliminating recording surge - Google Patents

Abstract

The invention relates to a method and a device for eliminating recording surge. The method for eliminating surge is applied to an electronic device which is provided with a microphone for recording voice to generate a voice signal, and comprises the steps of intercepting a signal section in which the surge occurs from the voice signal, carrying out compensation operation on a preset compensation signal and the signal section to eliminate the surge on the signal section, and then combining the compensated signal section and the residual voice signal of the intercepted signal section to form a complete voice signal. The device for eliminating the recording surge comprises: a signal intercepting unit, an amplifying unit, a compensating unit and a combining unit for combining the compensated signal section and the residual voice signal intercepted from the signal section to output a complete voice signal. Therefore, the authenticity and the integrity of the voice signal can be kept, and the improvement of the identification accuracy rate when the electronic equipment carries out voice identification is facilitated.

Description

Method and device for eliminating recording surge

technical field

The present invention relates to a method and device for eliminating recording surges, in particular to a method and device for eliminating recording surges that are generated on voice signals recorded by electronic equipment at the beginning of recording.

Background technique

With the continuous evolution and improvement of mobile (ie, mobile, hereinafter referred to as mobile) computing technology, it has become an important application trend to add the voice recognition function to portable electronic devices such as PDAs and mobile phones.

Please refer to FIG. 1 , which is a schematic block diagram of an internal circuit of an existing conventional electronic device with a speech recognition function. The existing portable electronic device (such as PDA, mobile navigator or mobile phone, etc.) 1 with voice recognition function in the past will directly send the voice signal (such as a voice command) recorded by its microphone 10 to a voice recognition system at its back end. The device 11 performs speech recognition, and sends the speech signal that completes the speech recognition to the processing unit 12 at the back end, so that it can execute the corresponding function according to the recognized speech command.

But the microphone 10 that is generally arranged on the portable electronic device has different electrical characteristics with the traditional independent microphone, especially when the microphone 10 is just started to record the moment (during about 0.7 seconds before), it will make during this period The recorded signal waveform has a huge offset, as shown in Figure 2, which is a waveform diagram of a speech signal with a surge. This phenomenon is usually called a rising bias. Two major effects. First of all, when there is no voice at the position where the surge occurs, the speech recognizer 11 will mistakenly identify the surge as a voice signal and cause the speech recognizer 11 to misjudge the speech segment; When generated, the surge will be superimposed on the voice signal, destroying the characteristics of the voice signal, distorting the voice signal, and causing the voice recognizer 11 to misjudge the correct voice. Both of these will have a huge impact on the accuracy of speech recognition, and thus reduce the user's trust and intimacy with the speech recognition system.

Therefore, in the past, a method of eliminating surges prohibited sending the signal recorded at the moment when the microphone was activated (for example, the first 0.7 seconds) to the speech recognizer, that is, the signal segment of the first 0.7 seconds in the voice signal was cut off and then sent to the The voice recognizer performs voice recognition, so as to eliminate the influence of the surge on the voice recognition. However, this method will cut off part of the voice segment together with the surge for the situation where there is voice input at the moment when the microphone is turned on, and still cause the back-end voice recognizer to misjudge the incomplete voice signal .

This shows that the above-mentioned existing method for eliminating recording surge and its device obviously still have inconvenience and defects in method, product structure and use, and need to be further improved urgently. In order to solve the problems of the method and device for eliminating the recording surge, the relevant manufacturers have tried their best to find a solution, but no suitable design has been developed for a long time, and there is no suitable general method and device. The method and structure can solve the above-mentioned problems, and this is obviously a problem that relevant industry players are eager to solve. Therefore, how to create a new method and device for eliminating the recording surge has become a goal that the current industry needs to improve.

In view of the defects in the above-mentioned existing method for eliminating recording surges and the devices thereof, the inventor has been engaged in the design and manufacture of this type of product for many years with rich practical experience and professional knowledge, and cooperated with the application of theories to actively research and innovate, in order to A new method and device for eliminating recording surge can be created, which can improve the general existing method and device, making it more practical. Through continuous research, design, and after repeated trials and improvements, the present invention with practical value is finally created.

Contents of the invention

The object of the present invention is to overcome the defects of existing methods and devices for eliminating recording surges, and provide a method and device for eliminating recording surges that can effectively reduce the speech recognition error rate of portable electronic devices. Therefore, it is more suitable for practical use.

The purpose of the present invention and the solution to its technical problems are achieved by adopting the following technical solutions. A method for eliminating recording surge according to the present invention is applied to an electronic device, and the electronic device has a microphone for recording voice to generate a voice signal. The method includes the following steps: (A), Intercepting a signal segment that produces a surge in the voice signal; (B), performing a compensation operation on a preset compensation signal and the signal segment to eliminate the surge; and (C), combining the compensated signal segment The signal segment and a remaining voice signal of the signal segment intercepted through the step (A) to form a complete voice signal.

The purpose of the present invention and the solution to its technical problems also adopt the following technical measures to further realize.

The aforementioned method for eliminating recording spikes, wherein in the step (A), the spikes occur within a predetermined time after the start of the voice signal, and the signal segment refers to the voice signal.

In the aforementioned method for eliminating recording spikes, in the step (B), the compensation signal is a spike signal with the same length as the signal segment and the same surge amplitude and position.

In the aforementioned method for eliminating recording spikes, in the step (B), the compensation signal is a spike signal with the same length as the signal segment, the same spike position and amplitude, but opposite phase.

The aforementioned method for eliminating recording spikes, wherein the compensation signal is to record multiple non-speech signals that generate spikes by repeatedly activating the microphone in advance, and then extract them from the multiple non-speech signals through step (A) in advance After the complex signal segments of the surge are generated, the complex signal segments are summed and averaged, and then multiplied by an amplification gain.

In the aforementioned method for eliminating recording spikes, in the step (B), the spikes on the signal segment are eliminated by subtracting the signal segment from the compensation signal.

In the aforementioned method for eliminating recording spikes, in the step (B), the spikes on the signal segment are eliminated by adding the signal segment to the compensation signal.

The purpose of the present invention and the solution to its technical problems are also achieved by the following technical solutions. A device for eliminating recording surges according to the present invention is provided at a microphone output end of an electronic device to receive a voice signal recorded by the microphone, and the device includes: a signal intercepting unit for the Intercepting a signal segment that generates a surge in the voice signal; an amplifying unit that amplifies a preset compensation signal according to the gain of the microphone; a compensation unit that performs a compensation operation on the signal segment and the amplified compensation signal, to eliminate the surge in the signal segment; and a combining unit for combining the compensated signal segment and a residual voice signal of the intercepted signal segment to output a complete voice signal.

The purpose of the present invention and the solution to its technical problems also adopt the following technical measures to further realize.

In the aforementioned device for eliminating recording spikes, the amplified compensation signal is a surge signal with the same length as the signal segment and the same surge amplitude and position.

The aforementioned device for eliminating recording surge, wherein the device for eliminating recording surge further includes a surge vector database, and the compensation signal is pre-stored in the surge vector database.

Compared with the prior art, the present invention has obvious advantages and beneficial effects. As can be seen from above technical scheme, main technical content of the present invention is as follows:

In order to achieve the above object, the present invention provides a method for eliminating recording surge, which is applied on an electronic device, and the electronic device has a microphone for recording voice to generate a voice signal, the method comprising: (A) Intercepting a signal segment that generates a spike from the voice signal; (B) performing a compensation operation on the signal segment with a preset compensation signal to eliminate the spike; and (C) combining the compensated signal segment and the remaining voice signal of the intercepted signal segment after step (A) to form a complete voice signal.

The burst occurs within a predetermined time after the start of the voice signal, and the signal segment refers to the voice signal within the predetermined time.

The compensation signal is a surge signal with the same length as the signal section and the same surge amplitude and position.

The compensation signal is a surge signal with the same length as the signal section, the same surge position and amplitude but opposite phase.

The compensation signal is obtained by repeatedly activating the microphone in advance to record a plurality of non-speech signals that generate surges, and then extracting the multiple signal segments that generate surges from the non-speech signals through step (A) in advance, and then After the signal segments are summed and averaged, they are multiplied by an amplification gain.

In step (B), the glitch on the signal segment is eliminated by subtracting the signal segment from the compensation signal.

In step (B), the glitch on the signal segment is removed by adding the signal segment to the compensation signal.

In addition, in order to achieve the above object, the present invention also provides a recording surge elimination device for realizing the above method, which is installed at a microphone output end of an electronic device to receive a voice signal recorded by the microphone. The device includes a signal intercepting unit, an amplifying unit, a compensating unit and a combination unit. The signal intercepting unit is used for intercepting a signal segment generating a surge from the voice signal. The amplifying unit amplifies a preset compensation signal according to the gain of the microphone. The compensation unit performs compensation operation on the signal section and the amplified compensation signal to eliminate the surge in the signal section. The combining unit is used for combining the compensated signal segment and the remaining voice signal of the intercepted signal segment to output a complete voice signal.

The amplified compensation signal is a surge signal with the same length as the signal segment and the same surge amplitude and position.

The device for eliminating recording surge also includes a surge vector database, and the compensation signal is pre-stored in the surge vector database.

By means of the above technical solution, the method and device for eliminating recording spikes of the present invention have at least the following advantages: the present invention intercepts the segment signal where the surge occurs in the speech signal, and uses a pre-obtained compensation signal to match the segment signal Compensation calculations are performed to offset the surge in the segment signal and retain the voice signal in it, so that the part of the voice feature will not be damaged by the surge, and can effectively improve the recognition accuracy of the back-end voice recognition.

To sum up, the present invention provides a method and device for eliminating recording spikes that can effectively reduce the speech recognition error rate of portable electronic devices. The method for eliminating surge is applied to an electronic device, and the electronic device has a microphone for recording voice to generate a voice signal. The method includes intercepting a signal segment where a surge occurs from the voice signal, and using a The preset compensation signal and the signal segment are compensated to eliminate the surge on the signal segment, and then the compensated signal segment and the remaining voice signal of the intercepted signal segment are combined to form A complete voice signal; thereby, the authenticity and integrity of the voice signal can be maintained, and it is helpful to improve the recognition accuracy of the electronic device when performing voice recognition. It has the above-mentioned many advantages and practical value, no matter in method, product structure or function, it has been greatly improved, and it has made great progress in technology, and has produced easy-to-use and practical effects, and it is better than the existing recording elimination method. The surge method and its device have multiple enhanced effects, so it is more suitable for practical use, and it is a novel, progressive and practical new design.

The above description is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable , the following preferred embodiments are specifically cited below, and are described in detail as follows in conjunction with the accompanying drawings.

Description of drawings

FIG. 1 is a schematic block diagram of an internal circuit of a conventional conventional electronic device with a speech recognition function.

FIG. 2 is a waveform diagram of a speech signal in which a burst occurs.

FIG. 3 is a schematic circuit block diagram of an embodiment of the device for eliminating recording spikes of the present invention applied to an electronic device with a voice recognition function.

Fig. 4 is a detailed circuit block diagram of the device for eliminating recording surge in this embodiment.

FIG. 5 is a block diagram of a circuit for generating compensation signals in this embodiment.

FIG. 6 is a waveform diagram of a non-speech signal with spikes.

FIG. 7 is a waveform diagram of a speech signal after the surge is eliminated by applying this embodiment.

Detailed ways

In order to further explain the technical means and effects that the present invention takes to achieve the intended purpose of the invention, below in conjunction with the accompanying drawings and preferred embodiments, the specific implementation methods and methods of the method and device for eliminating recording surges proposed according to the present invention , step, structure, feature and effect thereof, detailed description is as follows.

See also shown in Fig. 3, it is an embodiment of the device for eliminating the recording surge of the present invention, the device 3 for eliminating the recording surge in this embodiment is to be arranged on a portable electronic device 2 (such as a mobile phone) with a voice recognition function , mobile navigator or personal digital assistant (PDA) and other mobile devices) between the microphone 20 and the speech recognizer 21 for speech recognition, the user can usually press a preset quick key on the electronic device 2 (Fig. not shown in the middle) to activate the microphone 20 to input voice (such as a voice command), and the microphone 20 records a voice command (usually a name, a noun or a sentence) from the outside and generates a voice signal correspondingly. Disabled until the hotkey is pressed again. The voice signal generated by the microphone 20 is then sent to the device 3 for eliminating the recording surge, and then sent to the speech recognizer 21 for the speech recognition operation after the surge elimination operation is performed. Wherein, the voice recognizer 21 includes an endpoint detector module (not shown in the figure), which directly receives an external voice signal and judges whether the recorded voice is a voice signal.

Please refer to FIG. 4 again, which is a detailed circuit block diagram of the device for eliminating recording surge in this embodiment. The device 3 for eliminating recording surges in this embodiment mainly includes a signal interception unit 31, an amplification unit 32, a compensation unit 33, a combination unit 34, and a surge vector database 37, wherein:

The signal intercepting unit 31 is connected to the microphone 20 for receiving a voice signal 23 recorded by the microphone, as shown in FIG. 2 . Moreover, due to the influence of the inherent electrical characteristics of the microphone 20, the microphone 20 will be fixed at the front position of the voice signal 23 (for example, during the first 0.7 seconds) at the moment when the microphone 20 is activated to generate a surge 24 as shown in FIG. 2 . Therefore, in order to eliminate the surge 24, according to the characteristic that the location of the surge does not change, the signal interception unit 31 will first sample the voice signal 23 when receiving the voice signal 23 from the microphone 20, so as to intercept the occurrence of the voice signal 23 The segment signal 25 of the burst (i.e. the signal during the previous 0.7 seconds), and in the present embodiment, it is to sample the voice signal 23 (sampling 11552 points, equivalent to the signal length of 0.7 seconds), and then the burst will occur The fragment signal 25 is sent to the compensation unit 33, and the remaining speech signal (ie, the signal after the sampling point 11552) 26 of the intercepted fragment signal 25 is sent to the combination unit 34 for temporary storage.

The amplifying unit 32 is a multiplier in this embodiment, which receives a gain value of the microphone 20 and a compensation signal 29 from a surge vector database 37 . Wherein, the gain value is a variable, which is obtained by reading the gain value set for the microphone 20 in the system chip when the microphone 20 is activated. The compensation signal 29 is obtained in advance, and its generation method is as follows:

As shown in FIG. 5 , it is a circuit block diagram for generating compensation signals in this embodiment. First by repeatedly starting the microphone 20 to record (collect) a plurality of non-speech signals 27 (about 20 to 30) that generate bursts (same as the burst position of the above-mentioned speech signal), as shown in Figure 6, it is a burst wave Then these non-voice signals 27 are sent to the signal intercepting unit 31, so that these non-voice signals 27 are sampled one by one (get 11552 points), to intercept the first 0.7 points of these non-voice signals Include the complex segment signal 28 of surge in the second, then these segment signals 28 are summed up and averaged through a summation and averaging circuit 35, and then sent to a curve modification unit 36 (optional, dispensable ), the compensation signal 29 can be obtained, and the compensation signal 29 is pre-stored in the surge vector database 37 . Therefore, the compensation signal 29 is actually a spike signal with the same length and the same spike position as the segment signal 25 .

Therefore, when the compensation signal 29 is sent to the amplifying unit 32, the amplifying unit 32 will generate the current microphone gain value (for example, gain=3) according to the current gain value of the microphone 20 (for example, gain=6) and the compensation signal 29. The compensation signal 29 is multiplied by an appropriate gain value (6/3=2) and its phase is reversed, so that the amplified compensation signal 29 ′ is the same as the surge amplitude of the segment signal 25 but the phase is opposite, and then the amplified compensation The signal 29 ′ is output to the compensation unit 33 .

And in this embodiment, the compensation unit 33 is actually an adder, so when the amplified compensation signal 29' is sent to the compensation unit 33 to be added to the segment signal 25, it can be combined with the surge in the segment signal 25 Addition cancels out, eliminating the glitches in segment signal 25 .

In addition, it is worth pointing out that the compensation unit 33 can also use a subtractor instead of an adder. At this time, the compensation signal 29 only needs to be amplified without inversion, and the amplified compensation signal is sent to the compensation unit. When the segment signal 25 in 33 is subtracted, the surge in the segment signal 25 will be eliminated because of subtraction from the compensation signal.

Then, the segment signal 25' after the surge is eliminated is combined with the remaining speech signal 26 of the intercepted segment signal 25 in the input combination unit 34 to form a complete speech signal 23' as shown in Figure 7 before being sent into the The subsequent voice recognition operation is performed in the voice recognizer 21 . Therefore, please refer to Fig. 7, which is a waveform diagram of the speech signal after the application of the present embodiment to eliminate the surge. As can be seen from Fig. 7, after the speech signal 23 is processed by the surge canceling device 3 of the present embodiment, it can indeed be eliminated. The surge on the voice signal 23' is eliminated, and the characteristics of the original recorded voice signal can still be retained.

Please also refer to FIG. 3 , the voice signal through the voice recognizer 21 for voice recognition will be sent to the back-end processing unit 22 to perform corresponding functions according to the recognized voice. And since the voice recognition process of the voice recognizer 21 is a conventional technology and not the technical feature of the present invention, it will not be described in detail here.

Please refer to Table 1 again, which is a comparison table of experimental data before and after this embodiment is not applied and this embodiment is applied.

Table 1

According to the experimental data shown in Table 1, it can be seen that in the situation where the surge problem has not been solved, the microphone repeatedly records 384 nouns (such as names) spoken by two people in turn, and the speech recognizer 21 can correctly identify 351 of them Nouns (wrong 33); and after eliminating the surge in the voice signal with the device 3 of the present embodiment, among the 384 nouns recorded by the microphone, the speech recognizer can correctly identify 358 nouns (wrong 26 nouns) ), it can be seen that this embodiment can correct the original error rate from 33 to 7, that is, it can reduce the recognition error rate by 7/33=21.21%.

In addition, it can also be seen from Table 1 that, in this embodiment, whether the compensation signal is subjected to the curve modification processing by the curve modification unit 36 or not, the effect achieved is the same.

It can be seen from the above description that in this embodiment, the segment signal 24 in which a surge occurs in the voice signal 23 is intercepted, and a pre-obtained compensation signal 29 is used to perform a compensation operation on the segment signal 25 to cancel the segment signal 25. surge and retain the voice signal therein, and then combine the compensated segment signal 25' with the remaining voice signal 26 of the intercepted segment signal 25 to form a complete voice signal, thereby not only eliminating the surge, but also A part of the voice signal in the segment signal is retained, and the part of the voice feature is not damaged by the surge, which can effectively improve the recognition accuracy rate when the back end performs voice recognition.

The above descriptions are only preferred embodiments of the present invention, and do not limit the present invention in any form. Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention.

Claims (10)

1. A method for eliminating recording surge, applied to an electronic device, which has a microphone for recording voice to generate a voice signal, characterized in that the method comprises the following steps: (A), intercepting a signal segment that generates a surge in the voice signal; (B), performing a compensation operation on a preset compensation signal and the signal segment to eliminate the surge; and (C) Combining the compensated signal segment with a remaining voice signal of the intercepted signal segment after step (A) to form a complete voice signal. 2. The method for eliminating recording spikes according to claim 1, wherein in the step (A), the spike occurs within a predetermined time after the start of the voice signal, and the signal area A segment refers to the voice signal within the predetermined time. 3. The method for eliminating recording spikes according to claim 1, characterized in that in the step (B), the compensation signal is a signal with the same length as the signal segment and the same surge amplitude and position burst signal. 4. The method for eliminating recording spikes according to claim 1, characterized in that in the step (B), the compensation signal is a signal that has the same length as the signal segment and the same surge position and amplitude but Pulse signals with opposite phases. 5. The method for eliminating recording surges according to claim 3 or 4, wherein the compensation signal is to record multiple non-speech signals that generate surges by repeatedly activating the microphone in advance, and then go through the steps of (A) After extracting the complex signal segments that generate the surge from the complex non-speech signals, the complex signal segments are summed and averaged, and then multiplied by an amplification gain. 6. The method for eliminating recording spike according to claim 3, characterized in that in the step (B), the signal segment on the signal segment is eliminated by subtracting the compensation signal from the signal segment. surge. 7. The method for eliminating recording spikes according to claim 4, characterized in that in the step (B), the signal segment on the signal segment is eliminated by adding the signal segment to the compensation signal. surge. 8. A device for eliminating recording surges, which is installed at a microphone output end of an electronic device for receiving a voice signal recorded by the microphone, characterized in that the device includes: A signal interception unit, used to intercept a signal segment generating a surge in the voice signal; an amplifying unit, which amplifies a preset compensation signal according to the gain of the microphone; a compensation unit, which performs a compensation operation on the signal segment and the amplified compensation signal, so as to eliminate the surge in the signal segment; and A combination unit is used to combine the compensated signal segment and a residual voice signal of the intercepted signal segment to output a complete voice signal. 9. The device for eliminating recording spikes according to claim 8, wherein the amplified compensation signal is a surge signal having the same length as the signal segment and the same amplitude and position of the spike. 10. The device for eliminating recording surge according to claim 8, further comprising a surge vector database, and the compensation signal is pre-stored in the surge vector database.

Images