JP2009157278A - Device for detecting audio signal feature and method for detecting feature - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and device for detecting feature suitable for a content including an audio signal such as voice and a music piece, with low processing load. <P>SOLUTION: The device includes: an information analysis section 440 for obtaining classification information of a stream from an input stream; an analysis control section 450 for specifying a method for analyzing the input stream from the classification information obtained by the information analysis section; a feature parameter conversion section 460 which analyzes the input stream based on the analyzing method specified by the analysis control section, and which converts it to the feature parameter to be specified as a detection item; a detection processing section 470 for detecting whether or not, the detection item which is the feature parameter specified by the feature parameter conversion section, is included in the input stream; and an indexing section 480 for holding a detection result of the detection item which is detected by the detection processing section, so as to be retrieved when playing back the input stream. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention is a content including video signals and audio signals such as audio and music, which are mainly composed of moving image information and audio signals and distributed by the public broadcasting or cable network supplied as surface waves or radio waves from broadcasting satellites. The present invention relates to a method and a detection apparatus for detecting a suitable feature detection process with a low processing load.

  TV receivers that receive and play back video information (video signals) and audio signals (audio signals), recording and playback devices (video recorders) and video imaging devices (video cameras) that can record and store received video and audio, etc. The spread and development of this is remarkable. In addition, even in personal computers, there are many cases where a function for receiving a television broadcast and reproducing video and audio is already provided as a standard specification.

  In such an environment in which various contents can be easily acquired, for example, Patent Document 1 shows a description regarding a music detection request for separating a music part and other parts included in a music program or content. .

In Patent Literature 1, when music detection is performed, if the data to be detected is an encoded stream, the sum component and the difference component of the L-channel signal and the R-channel signal in the frequency domain obtained in the decoding process are described. It has been proposed to calculate a power ratio based on this and determine a threshold value.
JP 2006-301134 A

  Patent Document 1 is a method specialized in music detection and is suitable for content such as a music program, but does not make sense as detection processing for other news programs or variety programs.

  In addition, based on this detection method, for example, when CM (commercial message, configuration independent of the main content included in commercial broadcast) is used for detection, analysis is performed on signals for two channels (ch). Since it is not necessary to perform processing, useless processing will occur.

  In addition, when it is obvious that a speech signal such as a news program is mainly (a relatively narrow band), it is not always necessary to perform analysis processing on the components of the entire band. Will end up.

  Further, in the case of a stream including a multi-channel signal, the processing load increases because a 2ch signal is obtained by downmixing after decoding all channel data. In general, in the case of a multi-channel signal, a lot of main information is contained in the front center channel or LR channel.

  In addition, even when considering only music detection, the frequency domain is assumed when an AAC (Advanced Audio Coding) standard, which is an audio encoding method generally used in broadcasting and storage media, is assumed. There are some optional processes for signals, and their treatment is not clear.

  An object of the present invention is to specify an appropriate detection process for content including audio signals such as voice and music according to the content classification information, and execute an analysis method according to the specified detection process Thus, a method and a detection apparatus for performing (detecting) a feature detection process suitable for content with a low processing load are provided.

  The present invention has been made on the basis of the above problems. An information analysis unit that acquires classification information of the stream from the input stream, and an input stream from the classification information acquired by the information analysis unit, an analysis level, an analysis band, and Analyzing the input stream based on an analysis control unit that identifies analysis of at least one of the analysis channels and the analysis method specified by the analysis control unit, and converts the input stream into a characteristic parameter that should be specified as a detection item A feature parameter conversion unit; a detection processing unit that detects whether or not a detection item that is a feature parameter specified by the feature parameter conversion unit is included in the input stream; and detection of a detection item detected by the detection processing unit A recording unit for recording the result so as to be searchable when the input stream is played back. There is provided an apparatus.

  According to the embodiment of the present invention, in order to determine an appropriate detection process according to the meta information of the encoded data to be detected and acquire a necessary signal according to the detection process, the analysis level and the analysis signal band Alternatively, detection processing can be performed at high speed with a minimum processing load by controlling the analysis target channel. As a result, even a relatively low-speed and inexpensive processor can perform processing, and the cost can be reduced.

  In particular, when the analysis target is multi-channel, it is possible to reduce an increase in processing load more than necessary, and it is possible to realize a flexible detection process at a higher speed and according to the processing capacity of the system.

  In general, in the case of a multi-channel signal, a lot of main information is contained in the front center channel or LR channel. Considering this characteristic, an appropriate detection item (music detection, CM detection, etc.) is determined according to the content, and only the corresponding channel signal (front center channel or front LR channel) is extracted and appropriate (analysis level) By performing the analysis (with a band), it is possible to realize a high-speed feature detection process without performing an extra decoding process.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 shows an example of a video recording / playback apparatus (video recorder) to which an embodiment of the present invention is applied. The embodiment of the present invention described below is not only applied to a video recording / reproducing apparatus, but also, for example, in a personal computer, a video camera, or a portable terminal device capable of reproducing a video program, video content, etc. Applicable. It can also be sold as a program that can be supplied to a portable terminal device, a personal computer, or a video game terminal device.

  A video recorder (video recording / playback apparatus) 1 shown in FIG. 1 includes a satellite digital TV broadcast provided via a broadcasting satellite or a communication satellite, a terrestrial digital broadcast and an analog TV broadcast provided by a ground wave (spatial wave), Alternatively, a tuner unit (TV tuner) 10 having a function of receiving video content and the like supplied via a cable network is provided, and an output from the tuner unit 10 is a video analog-to-digital converter (Video ADC) 14 and an audio. The signal is input to an (audio / music) system analog-to-digital converter (Audio ADC) 16. An input signal from the external input terminal (Aux) 12 is also input to the video analog-digital converter 14 and the audio (sound / music) analog-digital converter 16.

  The video stream digitized by the video analog-digital converter 14 and the audio stream digitized by the audio (sound / music) analog-digital converter 16 are input to an MPEG encoder 20. A digital stream (MPEG2-TS (TS is an abbreviation of Transport Stream), etc.) from the external digital input terminal 18 is sent to the MPEG encoder 20 via an interface (I / F, interface) 19 such as IEEE1394 (or HDMI). Entered.

  When the TV broadcast signal supplied to the tuner unit 10 is a digital signal such as MPEG2-TS, the digital stream from the tuner unit 10 is input to the MPEG encoder 20 as it is.

  The MPEG encoder 20 encodes the input stream into MPEG2-PS or encodes into MPEG4-AVC, except for outputting the input MPEG2-TS as it is (passing through). In the present invention, when a digitally encoded stream is recorded, it is assumed that feature detection and indexing are performed based on content information, and the above-described encoded stream or pass-through stream is recorded. It is processing to apply to.

  The stream data processed in the MPEG encoder 20 is temporarily buffered in a high-speed memory such as an SDRAM (Synchronous Dynamic Random Access Memory) 22.

  The stream data buffered in the SDRAM 22 and subjected to the predetermined processing is transferred to the HDD 104, the disk drive unit 24 or the memory slot 26 at a predetermined timing according to the contents.

  The HDD 104 includes an HD (hard disk) to which a capacity of HDD (that is, a hard disk drive, for example, 1 TB (1000 GB)) is given.

  The disk drive unit 24 is a disc-shaped recording medium, and stores data (optical (100 GB for reproduction only, 15 GB for recording) and optical disk 102 for DVD standard (4.5 GB)) on the HD (High Definition) DVD standard. Stream) can be recorded, and already recorded data (stream) can be reproduced.

  For example, a card memory 106 having a capacity of about 2 GB is inserted into the memory slot 26 and used.

  Stream data reproduced from the optical disk 100 or 102, HD (in the HDD 104) or the card memory 106 via the disk drive unit 24, HDD 104 or memory slot 26 is transferred to the MPEG decoder (MPEG Decoder) 30 via the SDRAM 22. Is done.

  The MPEG decoder 30 can decode MPEG2-TS, MPEG2-PS or MPEG4-AVC according to the transferred stream.

  Video data (MPEG2-TS or MPEG2-PS) decoded by the MPEG decoder 30 is converted into a standard or high-definition analog video signal by a video digital-analog converter (Video DAC) 32, and a video output (Video Out) terminal 36. The video is displayed on the display device 52 by connecting the video output terminal 36 and the display device (monitor device / display unit) 52.

  On the other hand, the audio data decoded by the MPEG decoder 30 is converted into an analog audio signal by an audio (sound / music) system digital-analog converter (Audio DAC) 34 and supplied to an audio (sound) output (Audio Out) terminal 38. Is done. Note that audio or music is reproduced by connecting the audio output terminal 38 and a speaker (which may be built in the display device 52 or may be independent).

  When the data supplied to the MPEG decoder 30 is MPEG2-TS, the data is supplied as it is to the digital output (Digital Out) terminal 39 via the interface 37 such as IEEE1394 (or HDMI).

  Further, the recording / reproducing apparatus (HD DVD recorder) shown in FIG. The main control block 40 functions as a stream parser. Although not shown, the main control block 40 includes an MPU (microprocessor) or a CPU (central processor), and stores an EEPROM 42, a work RAM 44, a timer 46, etc. for storing firmware and various control parameters. Comes with. Needless to say, the main control block 40 manages streams between the SDRAM 22 and the MPEG decoder 30 and is used for video recording and video reproduction.

  An audio signal feature detection unit 410 is also inserted or connected to the main control block 40.

  The audio signal feature detection unit 410 will be described in detail later with reference to FIG. 2, but a stream separation unit 420, a stream analysis unit 430, a meta information analysis unit 440, an adaptive analysis control unit 450, a feature parameter conversion unit 460, a detection A processing unit 470, an indexing unit 480, and the like are included.

  In the encoded stream, first, meta information is separated in the stream separation unit 420, and detection items to be described in detail later are determined using FIG. The determined detection items will be described later with reference to FIG. 7, and an analysis level and an analysis band for an encoded stream are associated with each other. Accordingly, in the present invention, an optimum stream analysis method is set based on the content, and a function of detecting features (of an audio signal) at high speed can be realized.

  More specifically, the stream separation unit 420 shown in FIG. 2 separates the received or input encoded stream into audio encoded data and its meta information.

  The meta information separated in the stream separation unit 420 is input to the meta information analysis unit 440, and (same) audio encoded data is supplied to the stream analysis unit 430. The meta information includes, for example, electronic program information (EPG) that is broadcast together with contents including audio data.

  The meta information analysis unit 440 analyzes the program classification information corresponding to the received stream from the electronic program information, and outputs the obtained program classification information to the adaptive analysis control unit 450.

  The adaptive analysis control unit 450 determines a method for performing stream analysis from the obtained program classification information, outputs the determination result as control information to the stream analysis unit 430, and outputs detection items to the detection processing unit 470.

  The stream analysis unit 430 analyzes the audio encoded stream input from the stream separation unit 420, and outputs information obtained by the analysis to the feature parameter conversion unit. In this analysis, FIG. 7 will be used to explain later, but to what level is the stream analyzed by the control information from the adaptive analysis control unit, what signal band is analyzed, what channel is analyzed Processing is performed based on information such as whether to do.

  The feature parameter conversion unit 460 converts information (mainly spectrum) obtained as a stream analysis result by the stream analysis unit 430 into a feature parameter suitable for a detection item (to be specified (extracted as a detection item)), and performs detection processing. To the unit 470. Note that the characteristic parameters to be converted include power information of each channel.

  In the detection processing unit 470, based on the feature parameter obtained from the stream analysis result in the stream analysis unit 430, it is determined whether or not the detection item instructed from the adaptive analysis control unit 450 is detected, and the detection result is displayed as an indexing unit. Output to 480.

  An example of the detection processing in the detection processing unit 470 is shown in FIG. 6. For example, CM (commercial message, content independent of main content included in commercial broadcast) detection, music detection, cheer detection, and corner sound detection , Etc.

  In the case of CM detection, silence below a certain level inserted before and after the CM is determined from the power information obtained as a characteristic parameter, and the silence interval is a general rule of CM length (15 seconds or 30 seconds). Etc.) and the beginning / end of the CM is detected.

  In music detection, the difference in signal between the L channel (ch) and the R channel (ch) of the received L channel (ch) is different between Lch and Rch, which is a characteristic of the musical sound signal (variation of LR, etc.). And determining the start / end of the music section.

  In cheer detection and corner sound detection, the corresponding section is detected based on the shape of the spectrum obtained as the characteristic parameter.

  Based on the detection result from the detection processing unit 470 and the corresponding time information, the indexing unit 480 stores which event occurred at which time as index information that can be used for the search processing. The event corresponds to the detected feature, such as the start / end of CM, the start / end of music, and the like. The indexing unit 480 records the above-described indexing information in a predetermined recording area such as a lead-in area, a header information recording area, or a TOC (Table of contents) unique to the recording medium, depending on the recording medium. Needless to say.

  FIG. 3 is a schematic block diagram illustrating an example of the configuration of the adaptive analysis control unit described with reference to FIG.

  Based on the program classification information from the meta-information analysis unit 440, the search method determination unit 452 draws a table in which a preset program classification and detection items suitable for the program classification are associated with each other (shown as a table). See) and determine the detection item.

  For example, a music section corresponds to a music program, a cheer section where a spectator excites the sports program, and a CM (commercial) section as a detection item. Note that the CM section is widely detected in, for example, a music program, a sports program, a variety program, or a movie as shown in FIG.

  Further, the analysis level determination unit 454 determines to what extent the stream should be analyzed based on the above-described detection items shown in FIG.

  Also, the analysis band determination unit 456 determines a band for stream analysis (a band for analyzing the stream is specified).

  Further, the analysis channel determination unit 458 determines a channel to be analyzed when the audio encoded stream is a multi-channel (ch) signal having more channels than two channels. This is because the required analysis level, bandwidth, and analysis channel differ depending on the detection item. As shown in FIG. 7, a preset detection item and an analysis process (level, bandwidth, or channel) suitable for the detection item are set. ) Can be determined by drawing a table associated with it (referring to the table).

  FIG. 4 shows an internal block diagram of the stream analysis unit.

  The stream analysis unit 430 is based on the MPEG-2 AAC (Advanced Audio Coding) standard, and includes a syntax analysis unit 432, an inverse quantization unit 434, a joint stereo (JS) unit 436, and a TNS (Temporal Noise Shaping). It has at least part 438. Note that an analysis channel determination unit 433 for determining the above-described analysis channel in the case of a multi-channel stream is positioned (inserted) between the syntax analysis unit 432 and the inverse quantization unit 434.

  The syntax analysis unit 432 extracts a quantization parameter, a scale factor (spectral scaling information), or a decoding parameter such as correlation information between channels from an audio encoded stream by Huffman decoding or the like according to the AAC standard, and performs inverse quantization. Output to the unit 434. The analysis so far corresponds to analysis level 1 (not shown in FIG. 7, but it is always analyzed for any stream), and basic encoding parameters such as the number of voice or music channels and the sampling frequency. Is obtained.

  Here, when the input encoded stream is a multi-channel signal, the decoding parameter of only the front center channel or the front LR channel is output to the inverse quantization unit 434 according to the analysis channel designation from the adaptive analysis control. By doing so, the subsequent processing such as inverse quantization is performed only for the limited number of channels without greatly losing the characteristics as audio information. Therefore, the processing load is greatly reduced.

  For stereo signals of 2ch or less, the decoding parameters such as the quantized spectrum are output to the inverse quantization unit 434 as they are.

  The inverse quantization unit 434 inversely quantizes the quantized spectrum (analyzed by the syntax analysis unit 432) based on the scale factor to obtain an original linear scale spectrum. The analysis so far corresponds to analysis level 2 and an inverse quantization spectrum is obtained. However, since the subsequent joint stereo processing is not performed, depending on the band, the spectrum is a mixture of Lch and Rch sum signals and Lch cases.

  The joint stereo (JS) unit 436 performs MS (Mid / Side) stereo or IS (intensity) stereo processing on the spectrum dequantized by the dequantization unit 434, thereby performing the original Lch signal and Rch signal. The spectrum divided into and is obtained. The analysis so far corresponds to analysis level 3, and the spectrum of each channel is obtained.

  A TNS (Temporal Noise Shaping) unit 438 is an optional process that makes it difficult to perceive noise by concentrating (shaping) a noise component in a section having a high signal level when viewed with a time domain signal at the time of encoding. That is, when viewed from the stream analysis process, the TNS unit 438 applies a synthesis filter for restoring the original spectrum (after the time frequency conversion as viewed from the encoder) from the spectrum before the TNS process. Correct.

  In the case of a decoder, a PCM signal is obtained (required) by IMDCT (Inverse Modified DCT) processing, which is frequency-time conversion, for the obtained spectrum, but stream analysis only handles signals in the frequency domain. So it is not necessary. The analysis until after the TNS process corresponds to the analysis level 4 and is a highly accurate spectrum immediately before the frequency time conversion.

  The above-described stream analysis controls up to which process (analysis level) is performed according to the above analysis level based on the detection item instructed from the adaptive analysis control unit 450. This means that, for example, in the case of CM detection, it may be determined whether or not there is silence, and the signal for each channel is unnecessary and the spectrum of analysis level 2 is sufficient, and the subsequent processing is unnecessary. ing.

  In the case of music detection, the power ratio of the sum component and the difference component of the Lch signal and the Rch signal is important, and therefore a spectrum of analysis level 3 or higher is required. On the other hand, TNS processing that is easily applied to pulse-like signals such as speech signals (speech) is used less frequently, and even if applied, the application band is limited. It is.

  On the other hand, in the case of cheer detection or corner detection, although it depends on the detection accuracy, characteristics including the shape of the detected spectrum are required, and an analysis level 4 signal that is a highly accurate spectrum is required.

  In this way, by calculating only the spectrum up to the minimum required analysis level according to the detection item, the processing load is reduced and the processing speed can be increased.

  The same applies to the analysis band. If the program is mainly a speech signal (speech) such as a news program, it is sufficient to set the analysis band to about 7 kHz as shown in FIG. It is not necessary to analyze up to 24 kHz band (in the case of 48 kHz sampling). On the other hand, for a wideband signal including music, it is preferable to analyze the entire signal band included in the encoded stream.

  The analysis channel is analysis channel control in the case of a multi-channel stream. For example, in the case of CM detection, information for one channel is sufficient because silence determination is the base as described above. The load can be reduced by analyzing only the (front) center channel having a smaller processing load.

  On the other hand, in the case of music detection, detection processing is often performed using signal characteristics (power ratio or the like) between channels as described above, and therefore analysis of the front LR channel is required.

  In the case of cheer detection, since the announcer's live speech is often included in the center channel, it is effective to analyze using the front LR channel in order to pick up cheers more effectively.

  FIG. 5 is a flowchart when the above-described series of audio signal detection methods are implemented as software.

  First, in the stream separation step, the received or input encoded stream is separated into audio encoded data and its meta information (S1).

  For example, electronic program information (EPG) is extracted from the meta information separated in step S1 in the meta information analysis step (S2).

  In the meta information analysis step (S2), when EPG (electronic program information) or the like is extracted, the attributes of the program that can be read from the EPG, for example, meta information such as news, movies, and music are shown in FIG. Detection items are specified (determined) based on a table in which such program classification information and detection items are associated (S3). Note that CM detection, which is a common detection item, is set when there is no meta information or when any necessary attribute information is included in the content.

  When the detection item is specified, in the analysis level determination step, the stream analysis is performed based on the table associating the detection item as shown in FIG. The analysis level at that time is specified (determined) (S4).

  Similarly to the analysis level determination, the analysis band determination step specifies (determines) the band to be analyzed from the detection items based on the table as shown in FIG. 7 (S5).

  Subsequently, an encoded stream analysis process is performed based on these determination results.

  First, in accordance with the AAC standard, a decoding parameter such as a quantized spectrum, a scale factor (spectral scaling information) and / or correlation information between channels is extracted from an audio encoded stream by Huffman decoding or the like (S6). It is determined whether or not the process can be terminated at "1" (S7).

  If it is determined in step S7 that the analysis level can be terminated at "level 1" (S7-YES), the information obtained as the stream analysis result is suitable for the detection item (specified (extracted) as the detection item). The feature parameter conversion to be converted into the feature parameter is executed (S15), and the detection process is executed based on the feature parameter obtained from the stream analysis result (S16), indicating whether or not the detection item has been detected. The detection result is indexed (S17).

  In step S7, if it is determined that the analysis level is not "level 1" (cannot be terminated at "level 1") (S7-NO), whether or not an analysis channel is specified (whether or not an analysis channel is specified) ) Is checked (S8), and if not specified (S8-NO), step S10 and subsequent steps are executed for all channels.

  On the other hand, when the analysis channel is designated (S8-YES), “inverse quantization” is adopted as a method of analyzing the stream only for the designated analysis channel, and the quantized spectrum is based on the scale factor. Inverse quantization is performed to obtain the original linear scale spectrum (S10). Thereafter, it is determined whether or not the analysis level can be ended at “level 2” (S11).

  If it is determined in step S11 that the analysis level can be ended at "level 2" (S11-YES), steps S15 to S17 described above are subsequently executed.

  If it is determined in step S11 that the analysis level is “level 2” that is insufficient (cannot be terminated at “level 2”) (S11—NO), “joint stereo (JS)” is adopted as a method for analyzing the stream. The inverse quantized spectrum is subjected to MS (Mid / Side) stereo processing and IS (intensity) stereo processing to obtain a spectrum divided into the original Lch signal and Rch signal (S12). Thereafter, it is determined whether or not the analysis level can be ended at “level 3” (S13).

  If it is determined in step S13 that the analysis level can be terminated at “level 3” (S13—YES), the previously described steps S15 to S17 are executed in the same manner.

  If it is determined in step S13 that the analysis level is “level 3”, but the analysis level is insufficient (cannot be terminated at “level 3”) (S13—NO), “TNS” is adopted as the method for analyzing the stream, “Level 4” is set (specified), and the characteristics including the detected spectrum shape (rough shape) are analyzed (S14).

  Thereafter, steps S15 to S17 described above are executed.

  In this way, by detecting the detection method during stream analysis and controlling the analysis channel in addition to the analysis level and analysis signal bandwidth, even multi-channel streams can be detected at high speed with the optimum processing load. It can be performed. That is, it is possible to reduce the load for determining an appropriate (necessary) detection process using the meta information of the encoded data to be detected and acquiring a necessary signal according to the detection process.

  In other words, according to the analysis level determination result determined in the analysis level determination step, each stream analysis process is skipped (the stream analysis process ends when an analysis level at which analysis can be completed is found). Only the analysis process suitable for the encoded stream can be realized.

  In addition, the analysis band and analysis channel for each stream analysis process are limited according to the analysis band determination result in the analysis band determination step and the analysis channel setting by the analysis channel, so the processing load is reduced and high-speed processing is possible It becomes.

  The series of processing described with reference to FIG. 5 may be MPU or CPU firmware (not shown) of the main control block (stream parser) 40, or may be provided as a program written in the EEPROM 42 in advance. Good.

  As described above, according to the embodiment of the present invention, in order to determine an appropriate detection process according to the meta information of the encoded data to be detected, and to acquire a necessary signal according to the detection process, By controlling the analysis level and the analysis signal band, detection processing can be performed at high speed with an optimum processing load. At this time, music can be detected by a relatively simple calculation, and the cost can be reduced.

  Even when the allowable processing load is limited, a flexible detection process corresponding to the processing load can be realized by controlling the analysis level according to the detection accuracy.

  It should be noted that the content of the present invention is not limited to the form described here, and it goes without saying that various other forms can be taken without departing from the spirit of the invention. In the present invention, a video recording / reproducing apparatus (video disc recorder) has been described as an example. However, a television apparatus or a personal computer (PC) incorporating a video recorder, or a unitized audio that can be externally added. It goes without saying that (content) playback devices are also included. It should be noted that the embodiments may be implemented by appropriately combining them as much as possible, or by deleting some of them, and in that case, various effects resulting from the combination or deletion can be obtained.

1 is a schematic diagram illustrating an example of a video recording / playback apparatus (video recorder) to which an embodiment of the present invention can be applied. The schematic block diagram which shows an example of a structure of the audio signal feature detection part which can be utilized as one Embodiment of this invention. The schematic block diagram which shows an example of a structure of the adaptive analysis control part used in the audio signal feature detection part which is one Embodiment of this invention. The schematic block diagram which shows an example of a structure of the stream analysis part used in the audio signal feature detection part which is one Embodiment of this invention. The flowchart explaining the example which implement | achieves the audio signal feature detection part which can be utilized as one Embodiment of this invention as software. Schematic which shows an example of matching with the program classification information used in the search method determination part shown in FIG. 3, and a detection item. FIG. 4 is a schematic diagram illustrating an example of correspondence between detection items used in an analysis level determination unit illustrated in FIG. 3 and an analysis level that is an index of how much detail should be analyzed.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Video recorder (video recording / reproducing apparatus), 10 ... Tuner part, 20 ... Encoder, 24 ... Disk drive unit, 26 ... Memory slot, 30 ... Decoder, 40 ... Main control block (stream parser), 50 ... On-screen display ( OSD) control unit, 52 ... display device, 100, 102 ... recording medium (DVD standard optical disc), 104 ... HDD (hard disk drive), 106 ... recording medium (card memory), 410 ... audio signal feature detection unit, 420 ... Decoder unit, 430 ... stream analysis unit, 432 ... syntax analysis unit, 434 ... inverse quantization unit, 436 ... joint stereo (identification) unit, 438 ... TNS () unit, 440 ... meta information analysis unit, 450 ... adaptive analysis control unit, 452 ... part, 454 ... part, 456 ... part, 460 ... feature Parameter conversion unit, 470 ... detection processing section, 480 ... Indexing (Indexing) unit.

Claims (13)

An information analysis unit that obtains classification information of the stream from the input stream;
An analysis control unit for specifying to analyze the input stream from at least one of the analysis level, the analysis band, and the analysis channel from the classification information acquired by the information analysis unit;
Analyzing the input stream based on the analysis method specified by the analysis control unit, and converting the feature parameter to a feature parameter to be specified as a detection item;
A detection processing unit that detects whether or not a detection item that is a feature parameter specified by the feature parameter conversion unit is included in the input stream;
A recording unit for recording the detection result of the detection item detected by the detection processing unit so as to be searchable during reproduction of the input stream;
An audio signal feature detection apparatus comprising:   The audio signal feature detection apparatus according to claim 1, wherein the analysis level executable by the analysis control unit includes an analysis level for changing an analysis signal band of encoded data.   The audio signal feature detection apparatus according to claim 1, wherein the analysis level executable by the analysis control unit includes an analysis level for changing an analysis signal band of encoded data and a channel of the analysis signal. Get the classification information of the stream from the input stream
A method for analyzing the input stream from at least one of an analysis level, an analysis band, and an analysis channel from the obtained classification information is specified.
Analyzes the input stream based on the specified analysis method, converts it into feature parameters to be identified as detection items,
Detect that the detected item that is the specified feature parameter is included in the input stream,
Keep the detected detection items searchable when playing the input stream
An audio signal feature detection method.   5. The analysis method can be specified by executing a plurality of analysis levels as analysis levels of encoded data, and setting an optimum processing load by changing the analysis level as necessary. Audio signal feature detection method.   6. The audio signal feature detection method according to claim 5, wherein the analysis method is specified by changing an analysis level, an analysis signal band, and an analysis channel of encoded data. A stream separation step of separating the encoded stream into audio encoded data and meta information thereof;
An information analysis step of extracting information of at least an audio signal from the content included in the stream from the separated meta information; and
Based on the specific information extracted,
Whether the detection item can be detected in the first level analysis with reference to at least the audio signal information of the content extracted by the information analysis step,
In the analysis of the second level following the first level, it is determined whether or not there is an analysis designation for a channel of the audio signal. If there is a channel analysis specification, the specified channel is analyzed and the detection item can be detected.
Whether the detection item can be detected in the analysis of the third level following the second level,
Whether it is necessary to detect the detection item by the analysis of the fourth level following the third level,
And detecting a feature of the audio signal,
Indexing the detected detection items so that they can be searched during playback of the input stream;
An audio signal feature detection program including at least.   8. The audio signal feature detection program according to claim 7, wherein the first level in the feature detection step includes at least a step of obtaining an encoding parameter applied when the stream is encoded.   8. The audio signal feature detection program according to claim 7, wherein the second level analysis in the feature detection step includes at least a step of dequantizing the quantized spectrum based on a scale factor to obtain a linear scale spectrum.   In the second level analysis, the presence / absence of analysis channel designation is checked. If there is an analysis channel designation, the quantized spectrum is inversely quantized based on the scale factor for the designated channel, and the linear scale spectrum is obtained. The audio signal feature detection program according to claim 9, further comprising:   In the second level analysis, whether or not an analysis channel is designated is checked. If no analysis channel is designated, the quantized spectrum is inversely quantized based on a scale factor for all channels to obtain a linear scale spectrum. The audio signal feature detection program according to claim 9, comprising at least a step.   8. The audio signal feature detection program according to claim 7, wherein the third level analysis in the feature detection step includes at least a step of obtaining correlation information between channels.   8. The audio signal feature detection program according to claim 7, wherein the fourth level analysis in the feature detection step includes at least a step of obtaining a characteristic including the shape of the detected spectrum.

Images