Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
Claims 1-6 are pending. Claim 1 is independent. Claims were amended by a preliminary amendment filed together with the filing of the Application.
This Application was published as U.S. 2024/0371394.
Apparent priority: 27 April 2021 (JP).
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-6 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims of the instant Application require amendments for correcting antecedent basis issues that have resulted in serious ambiguity of intent. Aside from the antecedent basis issues, the language is unclear likely due to translation.
Please amend to clarify the language in addition to addressing the antecedent basis issues.
In Claim 1, “the same song” has no antecedent basis and the defect makes all of the dependents that inherit this defect indefinite.
There are other instances, as shown below, where it is not clear if the Claim is referring back to a previously introduced element or is starting with a new element. When referring back please use the article “the.” When starting with a new element, change the name so there are not instances of the same term repeating with the article “a.”
Generate antecedent basis for a phrase before using it with a “the” later on, for example:
divide each piece of the acquired singing data into a plurality of temporal sections to obtain divided singing data;
generate a feature amount output model for inputting information based on singing data of the divided section and outputting a feature amount of the singing data of the section from the divided singing data through machine learning,
The Claim can refer to “the divided singing data” only after having first defined what “divided singing data” is or where it comes from.
1. A feature amount output model generation system configured to generate a feature amount output model for inputting information based on singing data which is time-series voice data relating to singing of a song and outputting a feature amount of the singing data, the system comprising circuitry configured to:
acquire singing data for each of a plurality of songs used to generate a feature amount output model;
divide each piece of the acquired singing data into a plurality of temporal sections; and
generate a feature amount output model for inputting information based on singing data of the divided section and outputting a feature amount of the singing data of the section from the divided singing data through machine learning,
wherein the circuitry performs machine learning according to criteria based on a distance between feature amounts of singing data relating to the same song and a distance between feature amounts of singing data relating to songs different from each other.
See other example suggestions below:
2. The feature amount output model generation system according to claim 1,
wherein the circuitry performs machine learning so that the distance between the feature amounts of singing data relating to the same song is shorter than the distance between the feature amounts of singing data relating to songs different from each other.
3. The feature amount output model generation system according to claim 1,
wherein the circuitry determines a section of the singing data to be used for machine learning on the basis of a distance between feature amounts which are output by [[a]] the feature amount output model in a process of generation.
4. The feature amount output model generation system according to claim 1,
wherein the circuitry acquires the singing data including data indicating a length of a time-series pitch.
5. The feature amount output model generation system according to claim 4,
wherein the circuitry converts data indicating [[a]] the length of [[a]] the time-series pitch included in the divided singing data into a word, to obtain a converted word which is a character string corresponding to the length of a pitch for each consecutive identical pitch, and
generates a feature amount output model for inputting information based on the converted word.
6. The feature amount output model generation system according to claim 2,
wherein the circuitry determines a section of the singing data to be used for machine learning on the basis of a distance between feature amounts which are output by [[a]] the feature amount output model in a process of generation.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-6 are rejected under 35 U.S.C. 103 as being unpatentable over Kyuma (U.S. 20090165633) in view of Jansson (U.S. 20230125789).
Regarding Claim 1, Kyuma teaches:
1. A feature amount output model generation system [Kyuma, Figure 3 shows the hardware components including the various circuitries, different CPU and GPUs, and input and output components as well as memory and storage components.]
configured to generate a feature amount output model for inputting information based on singing data [Kyuma does not “generate” a model but includes the “feature amount output model” of the Claim because it conducts “singing voice analysis P1” and outputs “singing voice parameters” / “feature amount of signing data.”]
which is time-series voice data relating to singing of a song and [Kyuma, speech and therefore song are time-series voice data based on the definition of what time-series is. (Wikipedia: “In mathematics, a time series is a series of data points indexed (or listed or graphed) in time order.” https://en.wikipedia.org/wiki/Time_series.) ]
outputting a feature amount of the singing data, [Kyuma, Figure 4, “singing voice” input at and “singing voice parameters” obtained from “singing voice analysis” at P1.]
the system comprising circuitry configured to:
acquire singing data for each of a plurality of songs used to generate a feature amount output model; [Kyuma, Figure 3, “microphone 36” and Figure 7, “singing voice (singing)” being input to the “singing voice analysis, P1” to generate the “singing voice parameter” and overall “singing voice analysis data, D5.” The singing voice corresponds to the other input to the system which is shown as “music piece data, D2.” Figure 30 shows that first “read data of selected music piece S42” and then for that particular music piece move to “voice obtaining processing S44” and “analysis processing S45.”]
divide each piece of the acquired singing data into a plurality of temporal sections; and [Kyuma, audio (speech or song or music) arrives in units of frames which are temporal section whose duration depends on the type of data and choice of the engineer. Here too: “[0197] The following will describe the singing voice analysis processing. FIG. 30 is a flow chart showing in detail the singing voice analysis processing shown at the step S26. It is noted that in FIG. 30, a processing loop of steps S43 to S46 is repeated for every one frame.”]
generate a feature amount output model for inputting information based on singing data of the divided section and outputting a feature amount of the singing data of the section from the divided singing data through machine learning, [Kyuma, Figure 4, “singing voice parameters” output of P1. Figure 7 shows the types of “singing voice parameters” which are measured and Figure 17 shows an example of values obtained for each of these parameters. See [0020]-[0212] confirming that the various parameters are obtained for frames of audio. Kyuma does not “generate” a model but includes the “feature amount output model” of the Claim because it conducts “singing voice analysis P1” and outputs “singing voice parameters” / “feature amount of signing data.”]
wherein the circuitry performs machine learning according to criteria based on a distance between feature amounts of singing data relating to the same song and a distance between feature amounts of singing data relating to songs different from each other. [Kyuma, Figure 4 shows that “genre name” (resulting from “type diagnosis” of P6) and “recommended music pieces” (resulting from “singing voice music piece correlation analysis “ of P4) are selected according to similarity/correlation/distance between the characteristics of the input “singing voice” with songs or pieces of music that are input at D2 or the genres input at D1. Figure 16 shows the correlation (distance) of various pieces of music from each particular genre; Figure 25 shows the calculation of the correlation/distance. The correlation/distance determines the genre of the piece of music. Figure 8 shows the correlation/distance of the singing voice with different genres. Figure 30 shows that the singing voice is analyzed (S46) and its genre correlation list is output (S48). Figure 32 shows selection of the genre having the highest correlation value S82 for a singing voice sample. Correlation is distance between parameters: “[0116] Then, the singing voice parameter and a music piece parameter stored in advance in the memory card 17 … are compared with each other. Here, the music piece parameter is generated in advance by analyzing music piece data. The music piece parameter indicates not only a characteristic of a music piece but also which singing voice parameter of a singing voice the music piece is suitable for. Thus, as a tendency of a value of the singing voice parameter is more similar to that of the music piece parameter, the music piece is determined to be more suitable for the singing voice. Such a similarity is determined, and a music piece suitable for the singing voice (a singing way, a characteristic of singing) of the player is searched for. In the embodiment, Pearson's product-moment correlation coefficient is used for determining a similarity. The search result is displayed as a "recommended music piece". Further, in the embodiment, a music genre suitable for the singing way of the player (a recommended genre) is also displayed. As a result, when the player finishes singing the music piece, for example, phrases, "A genre suitable for you is OOOO. A recommended music piece is …" are displayed.” ] (This limitation, while poorly translated, by reliance on the Specification and the goal of the instant Application which is stated in the section with the heading “Technical Problem” as recommending keys/songs to the singer that match his singing voice, is interpreted as matching the input singing voice of the singer to various genres and thus songs of those genres. Amend to clarify the intent if it differs from the interpretation.)
Kyuma does not include generating a model by machine learning.
Jansson teaches:
1. A feature amount output model generation system configured to generate a feature amount output model for inputting information based on singing data … [Jansson is directed to training and therefore “generating” a neural network system / “machine learning mode” which takes as input musical content including vocal content and determines the type/genre of the input. “A system, method and computer product for training a neural network system. The method comprises inputting an audio signal to the system to generate plural outputs f(X, Θ). The audio signal includes one or more of vocal content and/or musical instrument content, and each output f(X, Θ) corresponds to a respective one of the different content types….”Abstract.]
…
generate a feature amount output model for inputting information based on singing data of the divided section and outputting a feature amount of the singing data of the section from the divided singing data through machine learning, [Jansson generates the model: “[0094] In one example embodiment herein, the model herein can be trained using an ADAM optimizer. … Then, a Short Time Fourier Transform is computed with a window size of 1024 and a hop length of 768 frames, and patches of, e.g., 128 frames (roughly 11 seconds) are extracted, which then are fed as input and targets to the architecture 500. ...” “[0159] Mass storage device 1130 additionally stores a neural network system engine (such as, e.g., a U-Net network engine) 1188 that is trainable to predict an estimate vocal and/or instrumental component(s) of a mixed original signal, a comparing engine 1190 for comparing an output of the neural network system engine 1188 to a target instrumental or vocal signal to determine a loss, and a parameter adjustment engine 1194 for adapting one or more parameters of the neural network system engine 1188 to minimize the loss. A machine learning engine 1195 provides training data, and an attenuator/volume controller 1196 enables control of the volume of one or more tracks, including inverse proportional control of simultaneously played tracks.”]
wherein the circuitry performs machine learning according to criteria based on a distance between feature amounts of singing data relating to the same song and a distance between feature amounts of singing data relating to songs different from each other. [Jansson, Figure 7, showing the training/generation flowchart which includes a “loss minimized? 708” which is attempting to minimize the distance: “[0096] … In step 704 the loss calculator 612 employs a loss function to determine how much difference there is between the output f(X, Θ) and the target, which, in this case, is the target instrumental (i.e., the magnitude of the spectrogram of the track “B”)….” “[0159] Mass storage device 1130 additionally stores a neural network system engine (such as, e.g., a U-Net network engine) 1188 that is trainable to predict an estimate vocal and/or instrumental component(s) of a mixed original signal, a comparing engine 1190 for comparing an output of the neural network system engine 1188 to a target instrumental or vocal signal to determine a loss, and a parameter adjustment engine 1194 for adapting one or more parameters of the neural network system engine 1188 to minimize the loss. A machine learning engine 1195 provides training data, and an attenuator/volume controller 1196 enables control of the volume of one or more tracks, including inverse proportional control of simultaneously played tracks.”]
Kyuma and Jansson pertain to detection of content/genre of an input sound in the context of Karaoke and it would have been obvious to improve the older system of Kyuma with the more powerful and newer machine learning methods of Jansson. T (See Jansson: “[0006] Deep learning models have recently emerged as powerful alternatives to traditional methods. Notable examples include Reference [25] where a deep feed-forward network learns to estimate an ideal binary spectrogram mask that represents the spectrogram bins in which the vocal is more prominent than the accompaniment. In Reference [9], the authors employ a deep recurrent architecture to predict soft masks that are multiplied with the original signal to obtain the desired isolated source.” “[0215] A user operates the media playback device 1902 to enter an input 1652 specifying selection of a track (e.g., in one example, this may be a karaoke-enabled track in a karaoke environment, although this example is non-limiting)….” See also [0004] for use in Karaoke.) his combination falls under combining prior art elements according to known methods to yield predictable results or use of known technique to improve similar devices (methods, or products) in the same way. See MPEP 2141, KSR, 550 U.S. at 418, 82 USPQ2d at 1396.
Regarding Claim 2, Kyuma teaches:
2. The feature amount output model generation system according to claim 1,
wherein the circuitry performs machine learning so that the distance between feature amounts of singing data relating to the same song is shorter than the distance between feature amounts of singing data relating to songs different from each other. [Kyuma, Figure 4, where after singing voice music piece correlation analysis at P4, “nominated music pieces” are selected and the “recommended music pieces” are displayed at P5. The correlations/similarity/distances are calculated between the features obtained as “analysis data” and are a measure of distance/similarity. The higher the correlation and similarity, the lower the distance. Kyuma uses a threshold/predetermined distance to separate the recommended pieces from the other pieces. Pieces with correlation above a threshold are selected. “[0124] Subsequently, singing voice music piece correlation analysis (P4) is performed. In this analysis, the music piece analysis data (D3), the music piece genre correlation list (D4), the singing voice analysis data (D5), and the singing voice genre correlation list (D6) are inputted. Then, based on these data and lists, correlation values between the singing voice of the player and music pieces stored in the game apparatus 10 are calculated. Only correlation values which are equal to or larger than a predetermined value are extracted from the calculated values to produce a nominated music piece list (D7).”]
Kyuma does not include generating a model by machine learning. Jansson, as applied to Claim 1, was cited for the teaching of the performing of modeling and training of the model by machine-learning methods and the rationale for combination remains similar to that provided for Claim 1.
Regarding Claim 3, Kyuma teaches:
3. The feature amount output model generation system according to claim 1,
wherein the circuitry determines a section of singing data to be used for machine learning on the basis of a distance between feature amounts which are output by a feature amount output model in a process of generation. [Kyuma, Figure 4, in two places the similarity/distance/correlation is used to select either a type/genre of music that is suited to the singing voice (Figure 4, P6: “type diagnosis” which generates a “genre name”) or when the “recommended music pieces” are selected at P5. Both are on the basis of the parameters that are extracted/generated from the input “singing voice” or input “music piece data.” See Figure 4, P1: “singing voice analysis” generating “singing voice parameter” and P2: “music piece analysis” generating “music piece parameter.” “[0122] …In this processing, the above processing (an operation of the player) is performed, and a singing voice of the player is inputted to the microphone 36. Voice volume data and pitch data are produced from the singing voice, and singing voice analysis (P1) is performed based on these data. Then, as an analysis result, a singing voice parameter is outputted, and stored as singing voice analysis data (D5). The singing voice parameter is a parameter obtained by evaluating the singing voice of the player in view of strength, a musical interval sense, a rhythm, and the like. The singing voice parameter basically includes items common to those of the music piece parameter. The singing voice parameter will be described in detail later.” “[0120] More specifically, in the music piece analysis (P2), musical score data in the music piece data (D2) is inputted for performing later-described analysis processing. As an analysis result, the music piece analysis data (D3) and the music piece genre correlation list (D4) are outputted. In the music piece analysis data is stored a music piece parameter which indicates a musical interval sense, a rhythm, a vibrato, and the like of an analyzed music piece. In the music piece genre correlation list is stored music piece genre correlation data which indicates a similarity between a music piece and a genre. For example, for a music piece, 80 points and 50 points are stored for a genre of "rock" and a genre of "pop", respectively….”]
Kyuma does not include generating a model by machine learning. Jansson, as applied to Claim 1, was cited for the teaching of the performing of modeling and training of the model by machine-learning methods and the rationale for combination remains similar to that provided for Claim 1.
Regarding Claim 4, Kyuma teaches:
4. The feature amount output model generation system according to claim 1,
wherein the circuitry acquires singing data including data indicating a length of a time-series pitch. [Kyuma, Figure 4, “[0122] … Voice volume data and pitch data are produced from the singing voice, and singing voice analysis (P1) is performed based on these data….”]
Regarding Claim 5, Kyuma teaches:
5. The feature amount output model generation system according to claim 4,
wherein the circuitry converts data indicating a length of a pitch included in the divided singing data into a word which is a character string corresponding to the length of a pitch for each consecutive identical pitch, and generates a feature amount output model for inputting information based on the converted word. [Kyuma, teaches that the variation in pitch is determined which means that duration/length of time during which the pitch is constant would be reflected in the variation in the pitch. “[0016] In a fifth aspect of the present invention based on the first aspect, each of the plurality of singing characteristic parameters and the plurality of comparison parameters includes a value obtained by evaluating one of accuracy of pitch concerning the singing of the user, variation in pitch, a periodical input of voice, and a singing range.” Additionally, “periodical input of voice” also indicates pitch. These values are input to the comparison process to find the suitable genres. ] (Note that in this Claim the words are natural language textual words but rather refer to Figure 2 of the instant Application and “626262” and the like that are shown in this Figure: “[0035] … The feature amount output model generation unit 13 sets a character string in which a value indicating the pitch is continuously lined up by the calculated integer as a word of the consecutive identical pitch.” This is merely a method of presentation/coding of the pitch values.)
Regarding Claim 6, Kyuma teaches:
6. The feature amount output model generation system according to claim 2,
wherein the circuitry determines a section of singing data to be used for machine learning on the basis of a distance between feature amounts which are output by a feature amount output model in a process of generation. [Kyuma, Figure 4, in two places the similarity/distance/correlation is used to select either a type/genre of music that is suited to the singing voice (Figure 4, P6: “type diagnosis” which generates a “genre name”) or when the “recommended music pieces” are selected at P5. Obviously, higher similarity/correlation means shorter distance. “[0126] Further, type diagnosis (P6) using the singing voice genre correlation list (D6) as an input is performed. In this diagnosis, a genre having the highest correlation value is selected from the singing voice genre correlation data, and its genre name is outputted. The genre name is displayed as a result of the type diagnosis together with the recommended music piece.” “[0124] … Only correlation values which are equal to or larger than a predetermined value are extracted from the calculated values to produce a nominated music piece list (D7).”]
Kyuma does not include generating a model by machine learning. Jansson, as applied to Claim 1, was cited for the teaching of the performing of modeling and training of the model by machine-learning methods and the rationale for combination remains similar to that provided for Claim 1.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to FARIBA SIRJANI whose telephone number is (571)270-1499. The examiner can normally be reached 9 to 5, M-F.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Pierre Desir can be reached at 571-272-7799. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Fariba Sirjani/
Primary Examiner, Art Unit 2659