FI119343B - Method for signal processing and signal processing apparatus - Google Patents

Description

119343

Method for signal processing and signal processing apparatus

The present invention is directed to a method for processing a frequency band limited signal, wherein a signal in analog form 5 is digitized, dividing said frequency band into at least first and second subbands, and searching one or more blue signals in the second subband. The invention further relates to a signal processing device 10 having means for converting a frequency-limited signal in analog form to a digital one, means for dividing said frequency band into at least first and second subbands, means for searching one or more blue signals from the second subband signal; from the signal contained in the subband.

Methods are known for converting an analog audio signal to a digital one and storing it on a recording medium or transmitting it over a data transmission channel to the receiving device. The signal converted to digital is converted back into a V ·: analog audio signal during playback, so that it is audible. In the analogy / digital conversion, samples of the audio signal are taken at substantially regular intervals, * * * representing the instantaneous values of the audio signal:: ··· 25. Since the sampling interval is not infinitely small, sampling of the original information takes place. The sampling rate used in analog / digital conversion sets the upper limit for how high frequencies can be \, lm converted to digital. According to Nyquist's theorem, the sampling rate · * 30 hairs should be more than twice the maximum signal frequency to be converted.

··· • * • • · ·: ·. * For high quality audio signal digitization, frequencies of up to 40 kHz are used, e.g., for recording on CDs, the sampling frequency is essentially 44100 Hz, whereby the maximum frequency of the audio signal to be converted is used. is in theory about 22050 Hz, but in practice about 20,000 kHz.

2 119343 In addition to the sampling rate, the quality of the analog / digital conversion is affected by the resolution used in the conversion, i.e. the resolution. Accuracy indicates how many different sample values can be generated in the analog-to-digital conversion. For example, in a 16-bit conversion, there are 65536 different possibilities, which is considered sufficient for high-quality conversion of the audio signal to digital.

10 The problem with the digitalization of the audio signal is e.g. large number of bits. For example, the above numerical values have 1411200 bits (= 2 x 44100 x 16), i.e. 176,400 bytes in a 16-bit sample of a two-channel signal lasting one second. In this case, a large amount of storage / communication capacity is required in order to store the digital-converted audio signal on a storage medium and / or to transmit to the recipient via the communication channel.

Various compression methods (compression methods) have been developed to reduce the number of bits. Compression methods are either lossless or lossy. Lossless **: * In which methods the original digital information is restored by decompressing the compressed information. In lossy comp -:. * Ί '· * decompression of the compressed signal may not *: ··: reproduce the original digital information in exactly the same way.

· · · · · · · · · · · · · · · · · · · · · · · · Many audio signal compression methods use lossy compression because it provides a much more efficient compression than lossless compression. Differences decom- ··. ·. between a compressed, analog-to-analog signal, and the original signal .., when using lossy compression, generally * ·: ** is not very easily heard by the ear. However, despite the lossy compression * "" *, not enough compression will be achieved in all situations, which may interfere with the recording / listening of the audio signal.

• · • * ”One known audio lossy compression method is * · 35 so-called. MP3. This method is relatively effective for signals with good correlation properties. However, at high frequencies, where noise is predominant, MP3 compression and other known compression methods are quite inefficient. Thus, much of the binary information generated in the compression consists of a portion of noise. A typical bit stream in MP3 compression is, for example, about 128000 bits per second.

5

It is an object of the present invention to provide a more efficient method of compressing and processing a signal having a limited frequency bandwidth in digital form, and a signal processing device. The invention is based on the idea of dividing the 10 frequency bands to be processed into at least two subbands and analyzing and searching for any blue signals occurring in at least one subband. More specifically, the method of the present invention is essentially characterized by dividing the second subband into at least two subbands, and determining the energy of the audio signal contained in each subband used to represent the noise in the signal. The signal processing device according to the invention is essentially characterized in that it comprises means for dividing a second time band into two or more subbands, means for determining the energy contained in each of the 20 subbands by displaying the noise in the signal;

The present invention provides significant advantages over the prior art *: **: 25 solutions. With the honey of the invention; · *; the method can convert the analog signal in a certain frequency band into a digital format such that the binary information needed for digitization can be represented with a smaller number of bits than the prior art methods, thus significantly improving the compression efficiency of *: ** 30. For example, in the example situation described above * "* ·, where the sampling rate of the two-channel 16-bit samples is essentially 44100 Hz, the original signal \ can be displayed in the method according to the invention at about 64000 bits / s. - * · *: 35 six, whereby, for example, less information is needed to store audio information and the audio signal can be transmitted on a data transmission channel at a lower bit rate than in prior art systems.

The invention will now be described in more detail with reference to the accompanying drawings, in which Figure 1 illustrates a frequency spectrum of an audio signal, Figures 2a and 2b illustrate the audio signal of Figure 1 converted to subbands by a method according to a preferred embodiment of the invention. 4 shows an audio signal reconstructed from a signal generated by the method according to the invention, 20 · .. *: * Figure 5 shows a V · signal processing device according to a preferred embodiment of the invention, • t · • · · ·:: · shows an absolute frame used in a method according to a preferred embodiment of the invention. Figure 6b illustrates a differential frame for use in a method according to a preferred embodiment of the invention - *: ** 30 structures in a simplified manner, - · Figure 6c shows a method for use in a method according to a preferred embodiment of the invention framed '! bit stream, · · · * 35 Figure 7a shows the implementation of signal changes between successive frames, and 5 119343 Figure 7b shows the implementation of fast signal changes between successive frames.

In the following, the invention will be described using, by way of example of an analog signal, an audio signal having a frequency band limited to 0-16 kHz. However, it is to be understood that the numerical values presented herein are merely illustrative and not limiting examples of the invention. Figure 1 shows an example of such an audio signal. In a method according to a preferred embodiment of the invention, this frequency band is divided into two subbands: a first subband and a second subband. The splitting is preferably performed such that both subbands are substantially the same width, but other splitting criteria can also be applied. Fig. 2a shows a first subband of the signal thus distributed, according to Fig. 1 15, and Fig. 2b, respectively, shows a second subband of the signal according to Fig. 1.

Figure 5 illustrates a signal processing device 1 according to a preferred embodiment of the invention, preferably comprising at least a signal encoding block 2 of signal 2. The signal encoding block 2 performs a ·· *: 'conversion of an analog signal to a digital signal according to the method of the invention. In addition, the signal processing device 1 may comprise a 0 '· signal decoding block 3 in which the *: *: signal in digital form is converted back to an analog signal. An analog signal to be converted to a digital ·: ··: 25 format is, for example, an audio signal produced in one or; ** \ multiple microphones (not shown), which may include music, vocals, speech, etc. The analog signal is fed to an analog / digital converter 4 conversion to digital. Preferably, the sampling rate is at least '*: ** at least twice the upper limit of the analog signal, i.e., at least 32 kHz in the system according to this precursor' · * '·. Analog: **]: The signal from the digital / digital converter in digital format is: · applied to the first 5 and the second filter 6. In the first *! The filter 5 preferably has low pass filtering, wherein the output of the first 35 filters has a first subband portion of the digitized analog signal, in this example in the frequency range 0-8 kHz. The second filter 6, respectively, performs 6119343 preferably a high pass filter part of the digitized analog signal, in this example, in the frequency range 8-16 kHz Filters 5, 6 are preferably digital filters, but it is clear that the analog / digital-5 conversion can also be performed after the filtering step using analog filters as filters 5,6.

The signal at the output of the first filter 5 is applied to the decimation block 7. In the decimation block, the sample frequency is preferably halved, i.e. the signal is decimated by two. This is possible because the output signal of the first filter 5 does not contain substantially higher frequencies than the upper limit frequency of the first subband. This upper limit frequency is at most V * of the sampling frequency fs. The disinfected signal is led to a compression block 8 where the signal 15 is compressed by any method known per se, such as MP3. It should be noted that the decimation block 7 is not required if the analog / digital conversion is performed only after the filtering step and for each subband separately, since then the sampling frequency can be selected according to the subband upper limit frequency.

20 ·; - The second subband includes mainly noise in the case of an audio signal, but occasionally v · 'harmonic frequencies or *: ** frequencies generated by a music synthesizer may also occur in this second subband. The output signal of the second filter 6 is applied to a first analysis block 9, where one or more, preferably up to four, blue signals are searched for. Figure 1 and Figure 2b, respectively, illustrate by way of example a single blue signal having a frequency of about 11.5 kHz. If such blue signals are found, the frequency and energy of these blue signals are investigated. Frequency and energy information **: * '30 is preferably stored in memory means 10. Subsequently, in block 11, the energies of the detected blue signals from the second subband signal are subtracted. Thereafter, the output of the separation block 11 is predominantly white noise.

After the difference block 11, the signal (noise signal) remaining from the signal of the second subband can be represented in a compressed form using the following method. In the distribution block 12, the signal 7 119343 of the second subband is divided into subbands, preferably eight subbands. These subbands are preferably substantially the same width, i.e. 1 kHz in this example. The second analyzing block 13 determines the energy of the signal contained in each subband.

5

Following the above steps, the various sections representing the digitized analog signal are combined in the bit stream generating block 14. This is preferably accomplished by forming frames in which one section is for the bits of the first subband compressed signal in-10, another section is for the frequency of and for binary information describing energy, and one third portion for binary information representing signal energy for each subband. Once the frame has been formed, it can be transferred to further steps such as storage on a storage medium, such as a CD, MD, and / or DVD. Frame information may also be transmitted to a communication channel, such as a telecommunications network, where the information may be transmitted to a recipient for conversion back into an analog signal.

The above steps are repeated during the processing of the analog signal.

I · * ··· \ *: further describes the conversion of the digitized analog signal into an analog signal in the v *: method according to a preferred embodiment of the invention, in the signal decoding block 3. The binary signal is led: ··; 25 is detected in the bit stream splitter block 15, where the first subwoofer ··· is separated. the information bits of the band and the second subband.

... The bits of the first subband are decompressed in the decompression block 16 and the up sampling in the up sampling block 17. The output of the up sampling block is then a substantially digital signal corresponding to the signal of the first *: ··: subband.

··· • ♦ · # «./ The second subband binary information separates the energy sections of the subbands 18 into the bit processing block 18 and the portions representing the possible blue signals. The subband energy information is fed to the noise generator 19, where In addition, the noise generator 119343 δ sa 19 generates sinusoidal signals whose frequency and energy are indicated in the sinusoidal portion of the binary information of the second subband.

After reconstructing the first and second subband digital signals 5 from the received binary information, they can be combined in the summing block 20. The combined signal is then converted to analog in a digital / analog converter 21. The analog signal can be output to a speaker 22 for example.

10

The digital signal information of the various subbands generated by the method of the invention can be combined in many different ways into a bit stream. This may be accomplished, for example, by using a frame structure according to the compression method used for compression of the first subband and adding portions thereof to the binary information of the second subband. Thereby, in some applications, it is possible to process the bit stream generated by the method according to the invention with a device for compressing a de-20 bit stream according to the compression method used in the first subband. For example, the second subband binary information added to the bitstream is interpreted as invalid frames. In this case, the decompressed analog signal comprises only a portion of the analog band signal's frequency band, but the sound quality may still be sufficient for the recording of music tracks recorded by the method of the invention: ···: 25.

· * · • · • ·

In the following, a bit stream generation according to a preferred embodiment of the invention for presenting signal information, e.g. for storage and / or transmission on a communication channel.

«· · *:: · Edullisesti 30 30 30 30 edullisesti::::::: 30 30 30 keh 30:::::::::::::::,,,,: Framing is preferably applied to bit stream generation, where ΓΊ different data can be represented in different parts of the frames (records). In this preferred embodiment, firstly, something as such is used emotionally; The frame structure of the prior art compression method, such as MP3, to represent information contained in the first subband. Preferably, two different types of frames are used to represent the second subband: absolute frame FR1 and differential frame 9 119343 FR2. Figure 6a shows the structure of absolute frame FR1. It comprises a header field H1 containing information on, among other things, the type of the frame, this type information being preferably not the reserved type of the compression method used, such a frame being not usually decompressed in the decompression step if the bit stream generated by the method of the invention is decoded The frame may still include a version field V1 to indicate which version of the frame information it is in. The frame still contains the length information L1 of the frame 10 and information about the signal itself. preferably provides subband number information SB1, i.e. information about which subband information is included in the frame (e.g. first five subband information). In addition, the frame possibly transmits information about the quantization table (not shown) used in the QT1 differential frames. Field TR1 indicates whether the frame is a normal frame or a so-called frame. transienttikehys. This means that the transient frame transmits information to the rapid change detected in the signal. This will be described a little later in this specification. In the case of a transient frame, information POS1 20 is still transmitted at which point the transient change is performed when decoding the signal. Field S1 indicates the number of detected blue signals * · (preferably between 0 and 4), so that the frame contains the corresponding amount of information • V * · about the frequencies SF1 and energy SE1 of the blue signals. Subbands Contains:: *: This noise transmits energy information NE1 for each of the available subbands *: ·> · 25. Based on the above frame information, de-. ·· *. in the coding step, it is possible to reconstruct the analog signal for this frame without previous knowledge of the signal.

* · * • · ·

The differential frame FR2 is somewhat similar to the absolute frame FR1 • »* ·; · * 30. The exception is that the frequency and energy data are not ·: ··: absolute values, but represent a change in these values: ***: to the corresponding values in the previous frame. In this way, the use of diffe- ♦ ·· ./ rentals frames can further reduce the signal **. the total amount of information to be generated. Also, the header · · · ** "35 field is not needed because in this embodiment it is assumed that the absolute frame is followed by at least one differential frame, although there may be a first 111143 frame between the frames, but in the decoder this is separated into its own bit stream.

Because errors may occur in the storage and transmission of data (bitvir-5 them) and, on the other hand, decoding may be started elsewhere than from the first frame, the number of successive difference frames FR2 is preferably limited. Thus, a certain number of differential frames FR2 is always followed by an absolute frame FR1. Therefore, in this preferred embodiment, the differential frame FR2 is provided with a counter which is incremented with each successive differential frame and set to its initial value (reset) in connection with the absolute frame. In this case, if during the decoding step it is found that the counter value is incorrect, an error may be assumed and the differential frame in question is not decoded. The decoding can be continued as the next 15 absolute frames become decodable.

The frames of the first subband and the frames of the second subband are preferably alternated in the transmitted bit stream. This is illustrated in Figure 6c. The first frame is an absolute frame FR1, followed by one or more differential frames FR2. One frame of the first subband is then transmitted and interleaving is repeated.

• · • · · ♦ · · V t

At the point when the signal-in-: ··: formatting transmitted in successive frames changes, the change cannot be directly transformed to the corresponding ·: ··· 25 changes in the audio signal, as this may be audible to the ear; ** \ annoying clicks, bangs, etc. the changes are made gradually from the starting value a to the target value b over a period of time: v. 7a. The change can be linear or non-linear: .. f linear. However, in some situations, the signal may contain relatively rapid changes, i.e. the transients mentioned earlier. In this case, *: **: the change is made much faster, for example in a quarter "of time compared to a normal change. The transient frame preferably contains information about where the change is made. This is illustrated in Figure 7b below. Notwithstanding the above, in some situations a substantially immediate change can also be made e.g.

a new blue signal has been detected, whereupon the frame in question is informed in any suitable way.

11 119343

Most of the functions of the present invention may be implemented, for example, in application software of a digital signal processing unit (not shown).

5

It will be understood that the present invention is not limited to the above embodiments, but may be modified within the scope of the appended claims.

· · · · · · · · · · · · · · · · · · · · · · · · · · · ··· · 1 · * · • · · · · · · · · · · · · · · · · · · · · · · · · · · ······ ···

Claims (11)

119343 The patent claimed the following: A method for processing a signal having a frequency bandwidth, wherein the signal in analog form is converted to digital, dividing said frequency band into at least first and second subbands and searching one or more blue signals in the second subband, substantially eliminating the effect of each detected blue signal. dividing the second sub-band into at least two subbands, and determining the energy of the audio signal contained in each sub-band used to represent the noise in the signal. 2. A method according to claim 1, characterized in that the first subband is formed by the low-pass filtering of the digitized signal, and that the second subband is formed by the high-pass filtering of the digitized signal. A method according to claim 1 or 2, characterized in that the signal of the first subband is compressed by any method known per se. A method according to claim 3, characterized in that the compressed signal of the first subband and the information generated from the second subband: T, representing one or more blue signals possibly found in the second subband and the energy of the audio signals contained in two or more subbands are transmitted. , interleaved over the communication channel. • · ·· «. . A method according to any one of claims 1 to 4, characterized in that frames formed from information formed from said subbands are formed. «· • · · • · · * ·. ** ·. The method of claim 5, characterized in that 35 absolute frames (FR1) and differential frames (FR2) are formed from information formed from said subbands, that information formed from said subbands is transmitted in absolute form in said absolute frames, and differential 119343 frames are used to indicate said subband changes in the generated values from the values of the previous frame. Method according to one of Claims 1 to 6, characterized in that the analog signal is an audio signal. A signal processing device (1) having means (2) for converting a signal of limited frequency bandwidth into an analog form to digital, means (5,6) for dividing said frequency band at least first and second openings, means (9) for transmitting one or more blue signals. searching means for the second subband signal, and means (11) for substantially removing the effect of each detected blue signal from the signal contained in the second subband, comprising means (12) for dividing the second subband into two or more sub-bands, means (13) for detecting noise in the signal by using the energies contained in said subbands. The signal processing device (1) according to claim 8, characterized in that the means (5, 6) for dividing said frequency band into at least the first and second subbands comprises a low pass filter (5) for forming a first subband and a second high pass filter (6). to form a subband. ··· * φ · • ♦ « The signal processing device (1) according to claim 8 or 9, characterized in that it comprises means (7) for decimating the signal of the first subband. • · 0 0 009 A signal processing device (1) according to claim 8, 9 or 10, characterized in that it comprises means (8) for compressing the signal of the first subband by any method known per se. . 0 · • · · • · • · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 8 w / w Wagner