AI Structured Summary
Not yet generated for this judgment
Judgment
Prabha Sridevan, J
The appellant claimed that it had invented ""A Chaos Theoretical Exponent Value Calculation system"" and applied for patent under
3624/DELNP/2005. The Deputy Controller held that the functions of the so-called system are based on mathematical method for solving mathematical
equations, and declined to accept the technical effect theory followed under European Patent law, as he was of the opinion that our law does not allow
patent for mathematical methods which have a technical effect, and therefore the invention was rejected as not patentable under S.3(k) of the Patents
Act 1970. The appellant has therefore preferred this appeal. The Invention relates to a system for analyzing a time series signal by a method based on
Chaos Theory and calculating a chaos theoretical exponent value (CTEV) thereof. The Background Art states that correlative dimension, KS entropy
and Lyapunov exponent are among the known chaos theoretical exponent values. The systems used to calculate the exponents include systems in
accordance with various procedures which are different algorithms (Wolf's, Kantz' Sano/Sawada and so on) Examples of prior arts have been
mentioned. And in particular the Complete specifications states ""in any of the systems, the Lyapunov exponents constitute the neighbourhood points
set generated from the neighbourhood condition to be set as the ratio with respect to its size on the strange attractor constructed in the embedding
space, and are calculated as the mean value when points constituting the neighbourhood points set separate from each other. The conventional chaos
theoretical exponent value calculation system uses one of such systems as mentioned above, and those systems have a presumption that it analyses a
system of stable dynamic (the dynamics is the behaviour limited by its physical form and the like or the property that provides the behaviour) (a system
of stable dynamics means a system with physically invariable disposition or length, and the shape of strange attractor generated from the time series
signal provided by the system becomes a similar for if such a system behaves chaotically). Thus the temporarily changing dynamics, or the Lyapunov
spectrum in Sano/Sawada's algorithm, cannot be calculated as a significant value (a system with its temporarily changing dynamics refers to a system
such as the human vocal organs, for example in which the physical disposition or length changes. For instance, when phonemes/a/and /or/ are
pronounced, the shapes of throat and oral cavity are different, and the strange attractor of the phoneme /a/ is enlarged or noises are added thereto).
For instance, in the analysis of a generic speech voice signal and the like, which is an exemplary system with a temporarily changing dynamics,
because a plurality of vowels change in a complex manner in a short period of time, the difficulty of analysis is extreme when compared with the
system using a conventional methodology. So far it is almost impossible to calculate a temporarily local first Lyapunov exponent in a system with a
temporarily changing dynamics such as an ordinary speech voice signal."" This also states the difficulties in the prior arts ""in the conventional methods,
it is not possible to make the temporal resolution to not more than 5 minutes, as well as it is limited to only quantize the mid- to long term stresses if the
reliability of index value is in only one significant digit, thus the real time or quasi real time evaluation of stresses in the speech of a speaker is
impossible. Thereafter a stable processing result cannot be obtained from a continuous speech voice (i.e. from a system with a temporarily changing
dynamics). The term ""stable"" means that the processing result does not almost very by a minute change of parameters.
The Summary of the Invention shows that what the inventors have invented is a system which makes it possible to calculate a CTEV that could not
have been so far processed in a dynamics-changing system and to perform the process thereof at a high speed and on a real time basis, and to
calculate a CTEV even from a time series signal which includes noises. The CTEV calculation system in accordance with the present invention is
referred to as SiCECA. ""In the SiCECA processing, since the sensitivity with respect to the noise intensity is adjustable, the chaos theoretical
exponent value calculation system has a higher reliability to allow the calculation of the intensity of chaoticity by setting the sensitivity to low, as well
as the calculation of ""a value obtained by adding to a value denoting the chaoticity intensity a value denoting the noise intensity"" by setting the
sensitivity to high, in addition to the calculation from the above of the intensity of the noise disturbing the chaoticity of the time series signal, and the
ratio of chaoticity intensity of the time series signal to the noise intensity.
Then it explains the invention according to each Claims. And then it says that ""From the experiments conducted heretofore, it has been made clear
that the chaos theoretical exponent value of a speech voice has a correlation with the degree of fatigue and the stress state of the speaker. The
present invention allows the measurement and evaluation of a psychosomatic state such as the degree of fatigue and the stress state of the speaker on
a real-time of quail real-time basis at the time of speaking.
The complete specifications compares the conventional techniques and states that the Invention calculates the average CTEV in a shorter time of
two decimal orders or more. The system in accordance with the present invention in contrast, which generates the neighbourhood points set for the
pro forma formal neighbourhood points set) from the periodicity condition, and which also allows the application of the neighbourhood condition or the
convergent calculation continuity condition in addition to the periodicity condition, makes it possible to calculate the chaos theoretical exponent value
when a stable dynamics is present and to obtain temporarily local processing result for more stable than ever.
It is effective to process the speech voice signal for the processing using the time series signal by the system in according with the present invention.
In the conventional techniques, the first Lyapunov exponent has been calculatable only when the clear voice signal by a single speaker can be
exclusively processed. The present invention allows the chaos theoretical exponent value to be calculatable on a real-time or quasi real-time basis,
which was previously impossible in practice, for example allowing the result to be displayed on a computer display if the time series signal is a speech
voice signal.
The chaos theoretical exponent value to be calculated by the chaos theoretical exponent value calculation system as mentioned above in according
with the present invention, in relation to the size of hypershere and the Lyapunov exponent calculated formally, is the value having the meaning as
shown in Fig. 7.
The brain is actively functioning, the speech voice includes many crosstalk noises, namely many disturbances of signals derived from other part, and
the relationship between the SiCECA neighborhood distance and the chaos theoretical exponent value calculated is on the curve of Fig. 7, indicating
the relationship between the chaos theoretical exponent value and the SiCECA neighbourhood distance parameter p is 10% is given by the vertical
line giving CEM10.
It can be seen from Fig. 7 that the sensitivity increases when the SICECA neighbourhood distance parameter is set to 10%, while the cerebral
evaluation sensitive decreases when 30%. At 100% the sensitivity for detecting any street and the like affecting the speech contents cannot be
obtained.
Since the dynamics that generated actual speech voice does not have deal noise characteristics as shown in Fig. 7, the SiCECA neighbourhood
distance parameter for obtaining a stable processing result must be set in correspondence with the degree of clarity of signals to be processed, and it is
better to set to 20% for the purpose of analyzing voices in an ordinary office environment.
In the description of the chaos theoretical exponent value calculation system (SiCECA) in accordance with the present invention, it is appreciated by
those skilled in the art how to use a memory, processor, storage means, etc., included in a computer in an ordinary fashion to calculate variables and
equations in the computer to express the procedure in an arbitrary programming language or a machine language and the like in order to execute on
the computer. For instance, variables and equations may be structured in arrays, and pointers and the like and may be expressed additionally using a
branch processing, a repetition processing, a re-entrant processing and the like.
For an ordinary speech voice, since the duration of one single phoneme is from tens milliseconds to one hundred and tens milliseconds, one
macroscopic chaos theoretical exponent value is one value calculated from hundreds to thousands microscopic chaos theoretical exponent values
calculated during that time, or a plurality of values in the order of several values depending on the processing parameter settings for the calculation of
a macroscopic chaos theoretical exponent value.
It is possible to obtain an exponent value showing changes over time of the cerebral activity and being visually understandable from a graph by setting
an average processing interval of about 30 seconds to perform a temporal moving averaging, for the macroscopic chaos theoretical exponent value
(S500).
The microscopic chaos theoretical exponent value as mentioned above means a chaos theoretical exponent value of a time series signal with respect
to each sampling time, and the macroscopic chaos theoretical exponent value means a chaos theoretical exponent value with respect to a
predetermined interval, such as the duration of a phoneme, based on the microscopic chaos theoretical exponent values.
Chaos theoretical exponent value.
The chaos theoretical exponent value calculation system in accordance with the present invention is a system for implementing such processes at a
computer terminal, which makes it possible to provide microscopic chaos theoretical exponent values for a system of a variable dynamics, which has
not been defined by any conventional system or to quantify the noise level of a c= noise that disturbs the chaoticity when convoluted on the chaos
theoretical time series signal, although the quantification of the noise did not have any special meaning in the conventional methods.
Accordingly, when constructing a system for observing any change in the cerebral activity of a speaker form his/her speech voice in accordance with
the present invention, it is preferable to provide a process for calculating temporal moving average exponents or an alternative process which makes it
easier to visually observe any change in the cerebral activity.
As have been described above, SiCECA is a system which makes it possible to calculate the chaos theoretical exponent values at a high-speed.
In the following description, the microscopic chaotic index value is represented by a micro-chaos theoretical exponent value Cm (= Cm1cro), and the
macroscopic chaos theoretical exponent value is represented by a macro-chaos theoretical exponent value is represented by a macro-chaos
theoretical exponent value.
CM (= CMACRO)'
First, the process (S200) for calculating the microscopic chaos theoretical exponent value will be described. To calculate the micro-chaos theoretical
exponent value Cm by SiCECA, an embedding dimension D, an embedding delay time Ëœd' an expansion delay time Ëœe' a size of neighbourhood
points set N are defined as chaos theoretical processing parameters, and the shortest period Tm and the longest period TM of the time series signal
are defined as the parameters of period settings in SiCECA (Sl10).
These parameters, an embedding delay time an expansion delay time te, the shortest period Tm and the longest period TM, will be defined using the
sampling interval At of the time series signal as a unit time for the sake of simplicity in the following description, hence td, te, Tm, and TM are used as
dimensionless numbers for representing the number of sampling intervals.
When the embedding delay time td, the expansion delay time te, the shortest period Tm and the longest period TM, require their time dimensions, these
parameters are explicitly defined as td x Δt te Δt, , Tm & Δt, and TM, Δt respectively.
The size of neighbourhood points set N is required to be (D + 1) or more of the embedding dimension D. In order to stably conduct the following
calculation without for example, zero-divide, it is preferable to set (D + 2), (D + 3) and more, however it is better not to increase it more than required,
to decrease the amount of calculations for the micro-chaos theoretical exponent values cm.
When the dynamics continuously varies such as in a speech voice, the size of neighbourhood points set N is preferably set as small as a stable
calculation is possible in order to prevent the intermixing of points of different dynamics into the neighbourhood points set or its candidate, from the
point of view of improving the reliability of the micro-chaos theoretical exponent values. For example, the appropriate size of neighbourhood points set
is 6 or 7 if the embedding dimension is 4.
Even when N is set to (D + 1), the use of a dithering, processing allows the prevention of the occurrence of zero - divide. For the size of
neighbourhood points set N, it is therefore important to set in accordance with the property of the signal to be processed in view of the improved
processing efficiency and the securing of reliability of the processing result.
After setting at 5110, because it is a calculation of the initial values at 5210 of Fig. 3, a time series signal used as input data for initial value calculation
is read (5220). Here the time series signal s = s(t), such as a continuous speech voice to be processed by. SiCECA, is defined as Equation 1.
Then a series of experiments are shown from Equation 1 to 29 and Equation 34.
The Complete Specification then deals with the Industrial Applicability
It says "" The microscopic chaos theoretical exponent value, in its form, is similar to the first Lyapunov exponent one of chaos theoretical exponents,
and SiCECA has partly a structure similar to the system using Sano/Sawada's algorithm as a system for calculating the Lyapunov exponents, in the
part which relates to aspect of micro-chaos theoretical exponent values.
However, SiCECA is a completely different system from the system using Sano/ Sawada's algorithm in its signal processing condition and procedure
as have been described above.
Sano/Sawada's algorithm corresponds to a generic time series signal processing, while SiCECA is applicable to signals having a periodicity, such as
speech voice signals. In addition, conventional systems including the system using Sano /Sawada's algorithm calculates, in principle, one single first
Lyapunov exponent for signal cut out as a processing unit, while in SiCECA, on the other hand, no explicit processing unit exists which is set to a
constant time duration or a fixed data size, and SiCECA calculates, in principle, a microscopic chaos theoretical exponent value for every sampling
time (or a microscopic chaos theoretical exponent value if the data to be processed is a speech voice.)
For example, in the system using conventional chaos theoretical signal processing algorithm such as Sano/Sawada's algorithm for using Lyapunov
exponents as exponents, the dynamics is assumed to be stable, thus these systems does not provide any effective results for the time series signal
from a system of changing dynamics.
The systems of conventional algorithms, also in the process of chaotic time series signal with a noise convoluted thereon, either vary initial parameters
in diverse ways in the processing or combine with another noise reduction system thereby to effectively calculate exponent values when no noise is
convoluted. However, there has been no system other than SiCECA, which quantifies the noise level itself in a precise way and with a processing
efficiency much higher in several decimal orders or more than the use of conventional systems.
SiCECA expands as Shiomi's Cerebral Exponent Calculation Algorithm. Shiomi is one of the inventors.
The learned Counsel for the appellant the following preliminary objections:-
a) The impugned order is incorrectly based on the claims filed along with the response to the FER on April 9, 2008 and does not take into account the
claims or the written submissions filed by the Appellant on July 25th 2008, subsequent to the hearing.
This letter dated July 25th 2008 is filed as Annexure 'L' in the paper book. This contains the written submissions from page 230 to 249. Page 250
deals with Amendment and it reads as follows:
To meet the objection of Section 3(k) of Indian Patent Act, we have made the following amendments in the claims:
* Substantially revised the claims
* Deleted the claims
* Added reference numerals in the claims
From Page 251 the amended claims start. As a sample the first page of this is extracted below:
We Claim:
A system for analyzing speech voice signal comprising:
A reading means (xi) for reading a speech voice signal to be subjected to a chaotic analysis;
A cutting means (xi) for cutting out said read speech voice signal for each processing unit for calculating a chaos theoretical exponent value of said
read speech voice signal, wherein said calculating means x(i) for calculating a chaos theoretical exponent value comprises:
A first calculation means for calculating a chaos theoretical exponent value with respect to said sampling time as a microscopic chaos theoretical
exponent value, in said cut-out speech voice signal at a processing unit; and a second calculation means for calculating the chaos theoretical exponent
value of said speech voice signal with respect to a predetermined time as a macroscopic chaos theoretical exponent value, based on said microscopic
chaos theoretical exponent value.
The system for analysing a speech voice signal as claimed in claim 1, comprising: A means for receiving, as parameters, an embedding dimension
D, an embedding delay time td, an expansion delay time te, a size of neighbourhood points set N, and the shortest period Tm and the longest period
TM of said speech voice signal;
Wherein said means for cutting out said speech voice signal for each processing unit cuts out a speech voice signal for each processing unit x=x(i)
from said speech voice signal based on Equation 2 as herein described, where, when said read speech voice signal is s=s(t), to and t1 in Equation 2 are
given as to and t1 satisfying a periodicity condition predetermined by Equation 3 as herein described
We called for the original records from the Patent Office. The letter is found there and also the written submissions.
But as far as the amendment is concerned, the tenor is different
Instead of the words ""we have made"" there is in writing ""we are ready to"" made. And instead of ""substantially revised"" it reads as ""to substantially
revise"", ""deleted"" reads as ""to delete"", and ""added"" reads as ""to add
These corrections are made in writing and initiated by the Counsel.
More importantly the amended claims are not found in the records.
We informed this to the Counsel for the appellant, who merely said that according to her instructions the amended claims were filed. We asked for
if she could furnish proof of filing the amended claims. She said it may be difficult. The documents filed in the Office on July 25th 2008 look complete
and clearly the amended claims were not before the Controller'. Therefore the submissions made by the counsel are at variance with what we have
found in the records.
This objection is rejected.
b) The next objection is that the Respondent did not substantiate the objection under S 3(k) till date of hearing where he mentioned that the subject
matter is objected to as it is a mathematical method.
We do not understand what the appellant means by saying that the Respondent did not substantiate. The Controller raises the objection and it is for
the applicant to substantiate his case. The Controller is not the adversary he is the authority who decides and his reasons will be found in the order.
Perhaps what the appellant means is that the Controller did not clearly say whether the objection is that it is a mathematical or business method or
computer programme per se or algorithms i.e. which of the kinds of invention barred under S.3(k) did his invention come under, in the opinion of the
Controller. Though on this ground we will not allow the appeal, it is always better if the Controller explains his objections clearly.
c) The next objection is that the Respondent had wrongly negated the 'technical effect"" holding incorrectly that the Indian Patent Law does not follow
the EPC.
This same ground was raised in Yahoo vs. Rediff and we had held:
When the patentee explains that there is an inventive step which is a technical advance compared to the existing knowledge (state-of the-art) or that it
has economic significance that would not give him the right to a patent as such. The inventive step' must be a feature which is not an excluded subject
itself. Otherwise, the patentee by citing economic significance or technical advance in relation to any of the excluded subjects can insist upon grant of
patent thereto. Therefore, this technical advance comparison, should be done with the subject matter of invention and it should be found it is not
related to any of the excluded subjects... In Symbian vs. Comptroller of Patents (supra), the application in question was ""Mapping dynamic link
libraries in a computing device"". It was a method of accessing data in a dynamic link library in a computing device. The Act in question was the UK
Act, 1977 and in S. 1(2) the Act excludes ""a scheme, rule or method... for doing business, or a program for a computer."" In Symbian the Court asked
the question Whether the claimed technical contribution can be said to be the excluded subject matter itself or whether the claim is actually technical.
That is why the UK and European Acts uses the phrase ""as such"" to limit the area of non-patentability. Perhaps Symbian acknowledges the lack of
clarity in the words ""as such"", and also that ""The danger is all the greater because the concept of a ""technical"" contribution is imprecise"". Symbian
approved of Merril Lynch's application which held that if there is ""some technical advance on the prior art in the form of a new result (e.g. A
substantial increase in processing speed as in Vicom)"" it might be possible to obtain a patent. Symbian also recognised the difficulty in being too
precise about deciding the feature. ""Each case must be determined by reference to its particular facts and features"" In Symbian, the Court dismissed
the Controller General's appeal against the order setting aside his refusal to grant patent on the ground of non-patentability.
The U.K. Court approved of the view of the Board in Gameaccount Ltd., T. 543/2006 where it was held that:
...It cannot have been the legislator's purpose and intent on the one hand to exclude from patent protection such subject matter, while on the other
hand awarding protection to a technical implementation thereof, where the only identifiable contribution of the claimed technical implementation to the
state of the art is the excluded subject-matter itself.
The Controller here was of the opinion that the invention which is the technical advance was itself nothing more than ""a mathematical method for
solving mathematical claims which are further based on various algorithms."" So the identifiable contribution was itself the excluded subject matter
according to the impugned order. So the Controller held that the Indian Patent law does not allow patent for a mathematical method just because it
provides a technical advance. His reasoning that merely because a mathematical method is a technical advance it cannot cross the 3(k) bar is right.
We see no reason to interfere with the impugned order. The appeal is dismissed.