CN103794218B - DVFS method for AAC - Google Patents
DVFS method for AAC Download PDFInfo
- Publication number
- CN103794218B CN103794218B CN201410039578.8A CN201410039578A CN103794218B CN 103794218 B CN103794218 B CN 103794218B CN 201410039578 A CN201410039578 A CN 201410039578A CN 103794218 B CN103794218 B CN 103794218B
- Authority
- CN
- China
- Prior art keywords
- frequency
- cpu
- stage
- aac
- dvfs
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The invention discloses a DVFS method for AAC. The DVFS method is applied to an embedded device with a CPU frequency-modulation pressure-regulating mechanism and includes the following steps of (1) reading audio stream information and coding specification information; (2) carrying out data coding at different stages; (3) realizing the frequency-modulation pressure-regulating mechanism. The sampling rate, sampling magnitude and the number of sound tracks in the audio stream information are obtained, overall complexity of AAC is determined by means of an AAC specification, and meanwhile, frequency and voltage of a CPU are dynamically regulated according to CPU real-time loads in the coding process by means of load variation characteristics of AAC. Accordingly, the purpose of reducing energy consumption is achieved.
Description
Technical field
The present invention relates under embedded system aac audio decoding techniques and in particular to a kind of for aac audio coding
Dvfs method.
Background technology
Embedded device, especially battery powered equipment, very sensitive to power problemses.Quickly send out in embedded cpu
Today of exhibition, increasingly stronger cpu performance also brings power problemses.In embedded soc power consumption, the power consumption of cpu occupies
Larger proportion, higher cpu power consumption not only makes the stand-by time of embedded device decline, more seriously, in circuit relatively
Big electric current makes stability and the service life reduction of cpu.Therefore, the high-performance soc chip of a new generation, much all has dynamic tune
Section cpu frequency and the function of voltage.According to cpu power consumption calculation formula e=cv2Ft, on the premise of meeting performance requirement, dynamically
Adjust the frequency of cpu and voltage (dynamic voltage and frequency scaling) becomes reduction cpu power consumption
A kind of effective means.
Aac coding (advanced audio coding) is a kind of file compression formats aiming at voice data design, with
Mp3 is different, and it employs brand-new algorithm and is encoded, and more efficiently, has higher " cost performance ".Aac can support many
Reach 48 tracks, 15 low frequency (lfe) tracks, 5.1 multichannels are supported, higher sample rate (reaches as high as 96khz, audio frequency cd
For 44.1khz) and higher sampling precision (supporting 8bit, 16bit, 24bit, 32bit, audio frequency cd is 16bit) and have many
The compatibility of kind of language, higher decoding efficiency, in general, aac can be on the premise of contrast mp3 file reduce 30%
More preferable tonequality is provided.
Traditional linux operating system has had the Policy model of dvfs, but these strategies are only the occupancy from cpu
Rate is set out, and arranges multiple FM thresholds and frequency modulation interval according to cpu occupancy, at set intervals, checks the occupancy of cpu,
According to whether reach threshold value to determine to be made whether to carry out raising frequency or frequency redution operation.The frequency modulation and voltage modulation precision of this dvfs strategy and
Speed is not high, not from the point of view of application, for specific aac coding application excessively rough it is impossible to enough reach standard
Really, the purpose of fast frequency hopping pressure regulation, accordingly, it would be desirable to for the feature of aac audio coding, customize a set of applicable dvfs method,
To reduce cpu power consumption further.
Content of the invention
Goal of the invention: reduce under embedded platform for effective, the power consumption when running aac coding application for the cpu, the present invention
A kind of dvfs method for aac audio coding is provided.
Technical scheme: to achieve these goals, the technical solution used in the present invention is:
A kind of dvfs method for aac audio coding, comprises the steps:
(1) audio stream and coding specification information are read；
(2) data encoding stage by stage；
(3) frequency modulation and voltage modulation mechanism is realized.
Described step () reads audio stream and coding specification information includes: obtains sample frequency rate of audio stream, adopts
Sample digit bits, channel number channels, determine the coding specification of aac encoder, and at present, aac encoder is mainly using lc
(low complexity) and two kinds of coding specifications of main.
Described step (two) before data encoding includes window selection stage by stage and window selection latter two stage: described
Cpu frequency according to required for the audio stream information obtaining and coding specification prediction for the last stage of window selection and corresponding electric
Pressure；Cpu according to required for the audio stream information obtaining, coding specification and window type are predicted for the stage after described window selection
Frequency and corresponding voltage.
After determining the coding specification adopting, encoder complexity is directly proportional to data volume size, rank before window selection
Cpu frequency freq required for section is directly proportional to sample frequency rate, sampling resolution bits, channel number channels, will encode
Specification as factor of influence x, that is, the cpu frequency freq=x × rate × bits required for the last stage of window selection ×
channels；Cpu frequency freq required for stage after window selection and sample frequency rate, sampling resolution bits, sound channel
Number channels are directly proportional, and the coding specification of employing is as factor of influence y, window type as factor of influence w, i.e. window selection
Frequency freq=y required for stage afterwards × w × rate × bits × channels.
In stage after window selection, window type includes long window and short window, and aac encoder in an encoding process can be to defeated
The undulatory property entering audio frequency is analyzed, and for steady-state signal, is encoded using long window, to obtain higher compression ratio, for
Transient signal, is encoded using short window, and to remove echo effect, when being encoded using long window, the load of cpu is more
Height, required frequency is bigger.
After obtaining cpu frequency freq, cpu can be obtained by the corresponding table of frequency-voltage inquiring about cpu and operate in this frequency
Required cpu voltage v under rate.
Described step (three) passes through application layer, drives layer and hardware layer to realize；Application layer runs aac coding application,
Dvfs decision-making point prediction next stage frequency, carries out frequency modulation and voltage modulation by driving layer operation hardware layer.
Application layer using the beginning of each frame cataloged procedure and carries out the decision point that window selection encodes dvfs as each frame,
In each dvfs decision point using the cpu frequency required for data encoding calculated next stage stage by stage, and will
Required frequency passes through ioctl function passes to driving layer；Drive layer to drive for hardware layer, be responsible for the access to hardware layer and control
System, and the interface function of frequency modulation and voltage modulation is provided to application layer；Hardware layer is responsible for the realization of frequency modulation and voltage modulation.
Described hardware layer includes a power management chip (pmic), can be adjusted rapidly by power management chip (pmic)
The frequency of section cpu and voltage.
Beneficial effect: the present invention, from application angle, is entering the sampling considering when line frequency selects in audio stream information
The influence factors such as rate, sample size, channel number stream information, coding specification, window type determine the overall complexity of aac coding,
It is the fine granularity dvfs scheme proposing for the application of specific aac audio coding, utilize the load variation characteristic of aac coding simultaneously,
In the cataloged procedure of each frame, after the beginning of a frame coding and window selection, two dvfs decision points are set, stage by stage
Prediction load, in an encoding process according to cpu real time load dynamic regulation cpu frequency and voltage, can significantly reduce cpu in fortune
Power consumption during row aac coding application.
With traditional the method for dvfs is carried out according to cpu occupancy compared with, this method has the advantage that
1st, frequency modulation and voltage modulation is more timely, and traditional dvfs method is carried out when cpu occupancy is detected above or below threshold value
Frequency modulation, and detect cpu occupancy time interval be generally 20ms～50ms, and for sample rate the voice data for 44khz,
The scramble time of one frame data be 22ms, therefore this dvfs method when for aac audio coding it is impossible to accomplish frame in and interframe
Timely frequency modulation and voltage modulation, and the present invention, after the beginning of a frame and window selection, is respectively provided with two dvfs decision points, thus
Reach the purpose of timely frequency modulation and voltage modulation.
2nd, frequency modulation and voltage modulation is more accurate, and traditional dvfs method determines usage frequency according to cpu occupancy, but aac compiles
Code does not only have the computing load of cpu, has substantial amounts of data accessing operation yet, is therefore only come really with the cpu occupancy in a certain moment
Fixed used cpu frequency is not accurate enough, and dvfs method proposed by the present invention has taken into full account the many of impact aac coding load
The factor of kind, and it is provided with a dvfs decision point so that frequency modulation and voltage modulation is more accurate in interframe and frame in.
In sum, for aac coding application, compared with traditional dvfs scheme, dvfs scheme proposed by the invention
Frequency modulation and voltage modulation is much sooner and accurately.The cpu lower power consumption 10% that aac can be encoded using the dvfs method that the present invention provides
～15%.
Brief description
Fig. 1 encodes flow chart for aac；
The system hardware and software structural representation that Fig. 2 realizes for frequency modulation and voltage modulation mechanism；
Fig. 3 is that frequency-voltage corresponds to table, and this figure is taking arm9 as a example；
Fig. 4 is frequency selective manner；
The dvfs method flowchart that Fig. 5 encodes for aac.
Specific embodiment
The present invention is explained further below in conjunction with the accompanying drawings.
For the dvfs method of aac coding, it is realized process and specifically includes that reading audio stream and coding specification information, divides
Phase data coding, frequency modulation and voltage modulation mechanism are realized.
Read audio stream and coding specification information: before carrying out data encoding coding, aac encoder obtains original wav
Or the information of pcm audio stream, information includes: sample frequency rate, sampling resolution bits, channel number channels；And determine and adopt
With which kind of coding specification (profile), at present, aac encoder is mainly using lc (low complexity) and two kinds of volumes of main
Code specification.
Data encoding stage by stage: the main flow of aac coding as shown in Figure 1, comprising: read audio stream information, determination
Coding parameter, allocation buffer, mental model prediction, window type selection, mdct conversion, instantaneous noise shaping, quantization, volume
Code.Aac coding be divided into two stages, wherein, window type select before be the first stage, determine adopt coding specification it
Afterwards, encoder complexity is directly proportional to data volume size, the cpu frequency therefore required for this stage and sample frequency, sampling resolution,
Channel number is directly proportional, using coding specification as factor of influence x, then the cpu frequency freq=x × rate required for the first stage
×bits×channels；Window type is second stage after selecting, and aac encoder in an encoding process can be to input audio frequency
Undulatory property be analyzed, for steady-state signal, encoded using long window, to obtain higher compression ratio, for transient state letter
Number, encoded using short window, to remove echo effect, when being encoded using long window, the load of cpu is higher, institute
Need frequency bigger, therefore by coding specification as factor of influence y, window type as factor of influence w, required for second stage
Frequency freq=y × w × rate × bits × channels.
Cpu frequency required for next stage is obtained by above-mentioned data encoding stage by stage, this frequency is referred to as ideal
Frequency corresponds to, it is generally the case that this ideal frequency is not the frequency shown in Fig. 3-voltage, the cpu frequency supported in table, and cpu can
To work at multiple frequencies, different frequencies is also different for the requirement of minimum supply voltage, and in general, cpu frequency is got over
Greatly, required supply voltage is also bigger.
Frequency modulation and voltage modulation mechanism realizes part: as shown in Fig. 2 by application layer, driving the cooperation of layer, hardware layer to realize frequency modulation
Pressure regulation mechanism.Wherein application layer is run and is arranged two at aac coding application, and the beginning in each frame and window type selection
Dvfs decision point；Drive layer to provide the interface function of frequency modulation and voltage modulation upwards for application layer, access downwards and control hardware layer；Hardware
Layer be generally a power management chip (pmic), by power management chip (pmic) can adjust rapidly cpu frequency and
Voltage.
Application layer using the beginning of each frame cataloged procedure and carries out the decision point that window selection encodes dvfs as each frame,
In each dvfs decision point using the cpu frequency required for data encoding calculated next stage stage by stage, now count
The frequency calculating gained is ideal frequency freqideal, it is generally between upper frequency freq corresponding in tablehighWith next frequency
Rate freqlowBetween, method proposed by the present invention is to allow cpu first operate in freqhigh, after reaching switch point, cpu
Operate in freqlow, until the task of current generation terminates, implement process as shown in figure 4, cpu runs the total of current task
Time is t, is t1 reaching the time before switch point, after switch point until the time that task terminates is
T2, then can determine switch point:freq by below equationideal× t=freqhigh×t1+freqlow×
T2, t=t1+t2.By the frequency selecting method shown in Fig. 4 in Fig. 3 shown in the corresponding table of frequency-voltage in select corresponding frequency
Rate, and required frequency is passed through ioctl function passes to driving layer.Driving layer is that power management chip (pmic) drives, and bears
Blame the access to power management chip (pmic) and control, and the interface function to application layer offer frequency modulation and voltage modulation, specifically
Frequency modulation and voltage modulation operation is completed by power management chip (pmic) chip.
It is as shown in Figure 5 that what the present invention was complete realizes flow process: first, before being encoded, obtains audio stream information (sampling
Frequency, sampling resolution, channel number) and coding specification information (lc or main)；Start setting up a decision-making what each frame encoded
According to the audio stream information obtaining and coding specification information, point, determines that aac encodes the cpu frequency required for the first stage, and presses
Enter line frequency according to the frequency selective manner shown in Fig. 4 to select；After window type selects, aac coding enters second stage, this
Cpu frequency according to required for the audio stream information obtaining, coding specification information and window type information determine for the stage, and according to
Frequency selective manner shown in Fig. 4 enters line frequency and selects；After aac coding second stage terminates, the cataloged procedure of a frame terminates,
Aac encoder reads next frame, and repeats said process.
The above be only the preferred embodiment of the present invention it should be pointed out that: for the ordinary skill people of the art
For member, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (6)
1. a kind of dvfs method for aac audio coding is it is characterised in that comprise the steps:
(1) sample frequency rate, sampling resolution bits, channel number channels and the coding specification information of audio stream are read；
(2) data encoding stage by stage: using the beginning of each frame cataloged procedure with carry out window selection and encode dvfs as each frame
Decision point, in each dvfs decision point using the cpu frequency required for data encoding calculated next stage stage by stage
According to the audio stream information obtaining and coding specification information, rate, determines that aac encodes the cpu frequency required for the first stage: window
Cpu frequency freq required for the last stage of selection is just become with sample frequency rate, sampling resolution bits, channel number channels
The coding specification of employing is as factor of influence x, i.e. cpu frequency freq=x × rate required for the last stage of window selection for ratio
×bits×channels；Cpu frequency freq required for stage after window selection, the coding specification of employing as impact because
Sub- y, window type as factor of influence w, that is, frequency freq=y required for the stage × w × rate after window selection ×
Bits × channels, frequency freq now calculating gained is ideal frequency, and it corresponds to table between frequency-voltage of cpu
Between a upper frequency and next frequency, cpu is allowed first to operate in, after reaching switch point, cpu operates in, until
The task of current generation terminates, cpu run current task total time be t, reach switch point before time be
It is t2 until the time that task terminates after t1, switch point, switch point be can determine by below equation:
freqideal× t=freqhigh×t1+freqlow× t2, t=t1+t2；
(3) frequency modulation and voltage modulation mechanism is realized.
2. the dvfs method for aac audio coding according to claim 1 it is characterised in that: described aac encoder
Coding specification includes lc and main.
3. the dvfs method for aac audio coding according to claim 2 it is characterised in that: described window selection it
In stage, window type includes long window and short window afterwards, and aac encoder can be carried out to the undulatory property of input audio frequency in an encoding process
Analysis, for steady-state signal, is encoded using long window；For transient signal, encoded using short window.
4. the dvfs method for aac audio coding according to claim 1 it is characterised in that: described step (three) lead to
Cross application layer, drive layer and hardware layer to realize；Application layer runs aac coding application, in dvfs decision-making point prediction next stage frequency
Rate, carries out frequency modulation and voltage modulation by driving layer operation hardware layer.
5. the dvfs method for aac audio coding according to claim 4 it is characterised in that: described application layer will be every
The beginning of one frame cataloged procedure and carry out window selection as each frame encode dvfs decision point, in each dvfs decision point
Using the cpu frequency required for the calculated next stage of data encoding stage by stage, and required frequency is passed through ioctl letter
Number passes to driving layer；Drive layer to drive for hardware layer, be responsible for the access to hardware layer and control, and provide tune to application layer
The interface function of frequency modulation pressure；Hardware layer is responsible for the realization of frequency modulation and voltage modulation.
6. the dvfs method for aac audio coding according to claim 4 or 5 it is characterised in that: described hardware layer bag
Include one for adjusting the frequency of cpu and the managing chip of voltage source.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410039578.8A CN103794218B (en) | 2014-01-27 | 2014-01-27 | DVFS method for AAC |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410039578.8A CN103794218B (en) | 2014-01-27 | 2014-01-27 | DVFS method for AAC |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103794218A CN103794218A (en) | 2014-05-14 |
| CN103794218B true CN103794218B (en) | 2017-02-01 |
Family
ID=50669805
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410039578.8A Expired - Fee Related CN103794218B (en) | 2014-01-27 | 2014-01-27 | DVFS method for AAC |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103794218B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101847413A (en) * | 2010-04-09 | 2010-09-29 | 北京航空航天大学 | Method for realizing digital audio encoding by using new psychoacoustic model and quick bit allocation |
| CN102184005A (en) * | 2011-06-03 | 2011-09-14 | 展讯通信（上海）有限公司 | Method and device for regulating dynamic voltages and frequencies |
| CN102446508A (en) * | 2010-10-11 | 2012-05-09 | 华为技术有限公司 | Voice audio uniform coding window type selection method and device |
| CN103021415A (en) * | 2012-12-12 | 2013-04-03 | 青岛天信通软件技术有限公司 | Digital-to-analog converter (DAC) voice data compression and uncompression technique |
| CN103345296A (en) * | 2013-06-04 | 2013-10-09 | 三星半导体(中国)研究开发有限公司 | Dynamic voltage frequency adjustment trigger device and method |
| CN103426453A (en) * | 2012-05-25 | 2013-12-04 | 华为技术有限公司 | Dynamic voltage frequency scaling method and system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7983204B2 (en) * | 2007-11-27 | 2011-07-19 | Motorola Mobility, Inc. | Wirelesss communication device and method |
| KR20130110459A (en) * | 2012-03-29 | 2013-10-10 | 삼성전자주식회사 | System on chip, electronic system having the same, and method for control of the soc |
-
2014
- 2014-01-27 CN CN201410039578.8A patent/CN103794218B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101847413A (en) * | 2010-04-09 | 2010-09-29 | 北京航空航天大学 | Method for realizing digital audio encoding by using new psychoacoustic model and quick bit allocation |
| CN102446508A (en) * | 2010-10-11 | 2012-05-09 | 华为技术有限公司 | Voice audio uniform coding window type selection method and device |
| CN102184005A (en) * | 2011-06-03 | 2011-09-14 | 展讯通信（上海）有限公司 | Method and device for regulating dynamic voltages and frequencies |
| CN103426453A (en) * | 2012-05-25 | 2013-12-04 | 华为技术有限公司 | Dynamic voltage frequency scaling method and system |
| CN103021415A (en) * | 2012-12-12 | 2013-04-03 | 青岛天信通软件技术有限公司 | Digital-to-analog converter (DAC) voice data compression and uncompression technique |
| CN103345296A (en) * | 2013-06-04 | 2013-10-09 | 三星半导体(中国)研究开发有限公司 | Dynamic voltage frequency adjustment trigger device and method |
Non-Patent Citations (4)
| Title |
|---|
| "AAC数字音频编码技术";陈英济;《电视论坛》;20051231;全文 * |
| "AAC音频编码中的比特分配技术";段筱雨;《声学技术》;20071031;第26卷(第5期);全文 * |
| "能耗期望的动态功耗管理预测性策略";黄少珉等;《应用科学学报》;20090531;第27卷(第3期);全文 * |
| "高级音频编码的一种信息隐藏方法";唐步天等;《声学技术》;20080831;第27卷(第4期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103794218A (en) | 2014-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2012246799B2 (en) | Method of quantizing linear predictive coding coefficients, sound encoding method, method of de-quantizing linear predictive coding coefficients, sound decoding method, and recording medium | |
| US8972252B2 (en) | Signal processing apparatus having voice activity detection unit and related signal processing methods | |
| US7711972B2 (en) | Pre-emptive power supply control system and method | |
| CN108847217A (en) | A kind of phonetic segmentation method, apparatus, computer equipment and storage medium | |
| CN102208188A (en) | Audio signal encoding-decoding method and device | |
| KR20120120085A (en) | Apparatus for quantizing linear predictive coding coefficients, sound encoding apparatus, apparatus for inverse quantizing linear predictive coding coefficients, sound decoding method, recoding medium and electronic device | |
| US9552845B2 (en) | Automatic generation of metadata for audio dominance effects | |
| CN101359978B (en) | Method for control of rate variant multi-mode wideband encoding rate | |
| CN101727906B (en) | Method and device for coding and decoding of high-frequency band signals | |
| CY1114289T1 (en) | LOW PROBLEM SOUND CONFIRMATION | |
| EP1747442A1 (en) | Selection of coding models for encoding an audio signal | |
| CN1828722A (en) | Complex signal activity detection for improved speech/noise classification of an audio signal | |
| CN103778918A (en) | Method and device for bit distribution of audio signal | |
| CN103544957A (en) | Method and device for bit distribution of sound signal | |
| CN110619884A (en) | Audio encoding method and apparatus | |
| CN103794218B (en) | DVFS method for AAC | |
| CN101924463A (en) | PFM control method of switching power converter and realizing device thereof | |
| WO2012139401A1 (en) | Audio coding method and device | |
| CN101425293B (en) | High-efficient sensing audio bit allocation method | |
| EP1943741B1 (en) | A device and a method of playing audio clips | |
| CN102779518B (en) | Coding method and system for dual-core coding mode | |
| US20050091052A1 (en) | Variable frequency decoding apparatus for efficient power management in a portable audio device | |
| CN101779111B (en) | Method and apparatus for encoding continuation sinusoid signal information of audio signal and method and apparatus for decoding same | |
| CN101388213B (en) | Preecho control method | |
| CN111968653B (en) | Light-load double-Bit self-adaptive incremental voice coding and decoding method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170201 Termination date: 20210127 |