Provisional Patent Application Serial 60/286,325, which in its entirety is incorporated by reference herein.

1. FIELD OF THE INVENTION This invention relates generally to toys and, more particularly, to talking dolls.

2. BACKGROUND Talking toys are not unique. There are many talking toys as shown by the number of talking dolls, vehicles, puppets, and animals now available. These talking toys, however, say the same preprogrammed sounds, words, phrases, or sentences, although the order in which they are spoken may vary. Furthermore, in most talking toys there is just one master voice chip and, hence, all the toys speak in that same voice or personality.

If a toy manufacturer desires to produce a product line that speaks two different languages, such as English and Spanish, the manufacturer creates two master voice chips, i. e. one chip recorded in English and the other recorded in Spanish. Talking toys that speak in Spanish contain the Spanish voice chip, while those that speak in English contain the English voice chip. These master

voice chips contain the recorded sounds, words, phrases, or sentences uttered by a professional voice talent (i. e. talented person). Typically, a professional talent is used rather than an engineer or an office staff. This voice chip is usually a memory device.

If a manufacturer desires to produce a line of talking toys that each say their own unusual or even unique respective combination or composition of name, birthday, lucky number, and favorite color, or that each say something different and in different families of voices, there is today no large-scale production process to create such toy lines at a cost low enough to be suitable and profitable for toys. The present invention fulfills this need.

SUMMARY OF THE INVENTION The invention relates to toys that say unique or apparently unique combination or composition of one or more speech elements. A toy, for example, may be able to say a speech such as"Hi ! My name is Sue Ellen and my birthday is October fifth. What's yours? My favorite color is burgundy, and my lucky number is five. Let's play !"This exemplary speech may be said in a number of master voices and is unique or apparently unique to the end-users, particularly children. Examples of master voices include Georgia Peach, Boston, Texas, Spanish,

Chinese, Indian, etc. Depending on the embodiment, a particular speech may be said by one or more toys in a product line, or by one or more toys in each master voice family.

Speech elements, in one embodiment, are digitally captured recordings, typically recorded by professional voice talents. They are stored in memory devices. Using a speech chip, the speech elements are assembled to form a combination of at least one speech element that is unique or apparently unique. In another embodiment, speech elements are stored as text. Using a speech chip with a text-to-speech synthesizer, the speech elements are assembled to form a combination of one or more speech elements. In another embodiment, the speech elements are digitally manipulated core master voices. Acoustic or sound characteristics of a core master voice are manipulated to create one or more new master voices.

In one embodiment, a serial number is used to indicate which combination of speech elements is unique or apparently unique. This serial number is then used to do a look-up to a dictionary of speech elements. In another embodiment, a serial number is not used, but rather the speech elements that are unique or apparently unique to the toy are stored in a memory device, so a look-up is unnecessary.

The invention also provides for a method of

programming these serial numbers into memory devices. It also provides for a method of programming these unique or apparently unique combinations/compositions of speech elements.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a conceptual elevation, mostly of the exterior but partly cut away to show certain interior electronic circuit components, of one preferred embodiment of a unique or an apparently unique talking toy in accordance with a preferred embodiment of the invention; Fig. 2 is a basic flow chart of exemplary operations in the Fig. 1 speech-engine circuit to cause the Fig. 1 talking toy to speak ; Fig. 3 is an exemplary electronic schematic of the Fig. 1 speech-engine circuit, including a speech chip, in accordance with a preferred embodiment of the invention; Fig. 4 is a high-level block diagram of the Fig. 3 speech chip in accordance with a preferred embodiment of the invention ; Fig. 5 is a representative high-level flow diagram for programming the Fig. 1 toy with a unique or apparently unique vocabulary in accordance with a preferred embodiment of the invention;

Fig. 6 is a representative high-level flow diagram for creating a line of unique or apparently unique talking toys ; Fig. 7 is a representative diagram for the assembly of talking dolls that utilize differing master voice families reflecting different regions and dialects, in accordance with preferred embodiments of the invention; Fig. 8 is a circuit diagram for each of multiple finished printed circuit boards (Fig. 3 speech engines), in accordance with a preferred embodiment of the invention ; Fig. 9 is a circuit diagram of a speech engine that includes a memory device containing an entire library of master voice families in a product line, in accordance with an embodiment of the invention ; Fig. 10 is a circuit diagram of a speech engine that includes a memory device but generally containing only unique or apparently unique speech elements, in accordance with an embodiment of the invention; Fig. 11 is a representative flow chart for the creation of multiple new subfamilies of voices from one core master voice; Fig. 12 is a variant of Fig. 8 but wherein the voice families are based upon one core voice family; and

Fig. 13 is a circuit diagram of a speech engine but including a text-to-speech synthesizer, in accordance with a preferred embodiment of the invention.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS A talking toy 100 (Fig. 1) of one preferred embodiment includes a toy (doll) 102, a printed circuit board (PCB) 112, a speech chip 108, a serial-number device (S/N device) 114, and a press-to-talk switch 106 to trigger the speech chip 108. The speech chip 108 preferably includes a memory, preferably a read-only memory (ROM) 110. The ROM 110 generally contains an operating program 116 and recorded speech elements 118 (both further discussed below). The operating program and recorded speech elements may also be stored in separate memories. The memory or memories 110 may also be external to the speech chip 108 so long as the speech chip 108 can interface with them.

In one embodiment, this talking toy 100 is uniquely (or very nearly so) different from other talking toys in the same product line because the toy has a unique or very infrequently occurring (apparently unique) combination/composition of name (first and middle), birth date, and master voice family (discussed further below).

This enables a toy manufacturer to produce a line of

virtually unique talking toys where there are no or very few toys from the same voice family (further discussed below) that have exactly the same name and birth date, somewhat analogously to real kids.

When the press-to-talk switch 106 is pushed, the speech chip 108 enables the doll 100 to utter a particular set of speech elements (e. g. sounds, music, words, phrases, or sentences) 104.

Examples of Doll Speech Showing Unique or Near- Unique Identities a) In a soft Georgia drawl kind of voice : Hi! My name is Sue Ellen and my birthday is October fifth. What's yours? My favorite color is burgundy, and my lucky number is five. Let's play! b) In a Yankee Boston kind of voice : Hello. My name is Jennifer Elizabeth and my birthday is April sixteenth.

My lucky number is seven, and my favorite color is canary. c) In a California Valley Girl kind of voice: Hi! My name is Bunny Sun and my birthday is June twentieth. What's yours ? My lucky number is two and my favorite color is turquoise.

d) In a Spanish maiden kind of voice (in all Spanish) Hola! My name is Maria Linda and my birthday is May nineteenth.

My lucky number is eight and my favorite color is sienna.

Master Voices In one embodiment of the invention, several master voice families (master voices or voice families) are used. These master voices reflect different personality types, regions, dialects, and/or countries. The table below is an example list of master voices.

English Language 1-Georgia Peach drawl 2-Boston Yankee 3-New York City"nue yawk" 4-Chicago nasal style ("she caah go") 5-Minnesota (Midwest) 6-Texas Girl drawl ("hi yew awl") 7-California Silicon Valley 8-California Valley Girl ("like hellllooo-there") 9-Canadian English ("eh my name is...") Other Languages 1-Spanish style 1 (Mexico dialect 1) 2-Spanish style 2 (South America dialect 2) 3-Spanish Style 3 (Spain dialect 3) 4-German 5-Italian 6-French 7-Japanese 8-Chinese (Mandarin dialect 1)

Each master voice typically is recorded by a different professional voice talent. In one embodiment, each talent records, in a unique dialect (or language) and character personality, one hundred twenty-one speech elements. These speech elements are digitally captured for recording 118 and retention in the memory 110 of the speech chip 108. In this way the various recorded speech elements are physically embodied in the ROM 110.

If a manufacturer, for example, desires to produce a line of talking dolls with four master voices, such as Georgia Peach drawl, Boston Yankee, New York City, and Chicago, the manufacturer preferably uses four electronic speech chips 108, each containing the recorded speech elements of the appropriate voice family 118. Using these speech chips 108, some of the talking dolls 100 are produced with speech elements recorded from the Georgia Peach voice 118, some from the Boston Yankee voice 118, and so on.

Speech Elements (Master Words)

In one embodiment, master voice recording sessions preferably require about one hundred twenty words to be spoken in character dialect and digitally captured. This may be accomplished in typically three to four hours of studio time per voice talent. The table below lists exemplary speech element types that may be recorded for each master voice family. Speech Element Word Count Exemplary Values Types (Master Words) to be Recorded Month name 12 words January, February, March,.. . December Month date 31 words First, second, thirteenth,.. . thirty-first Lucky Number 10 words One, two, three, nine, ten Favorite Color 8 words Turquoise, chartreuse, canary, burgundy, etc. Key Phrase 20 words Hi Hello ; My name is; My birthday is; My lucky number is My favorite color is What'syours? Let'splay! Name Element 40 words Sue, Ellen, Marie, Rosemarie, Frances, Nadia (the name elements may be different for each master voice family, e. g. Spanish names for Spanish-type master voice- Maria, Consuelo, Corazon, Solidad, etc.) TOTAL121 words

If we assume that it takes an average of one second to utter a word, the average time to record the one hundred twenty-one words above is one hundred twenty-one seconds. In this embodiment, the speech chip 108 has at least one hundred twenty-one seconds of speech memory 118.

Each speech element contains a value. This value is associated to a particular speech element type (e. g. month name, month date, first name element, middle name element, and the like) and to a particular master voice family (e. g. Georgia Peach, Boston, New York, and the like). Variations on the speech elements may also be done, such as using other speech element types (e. g. age, zodiac sign, music, and pet's name), using different values on the speech elements, having a different number of speech elements, and the like.

Name Elements (Names) Each of the talking dolls preferably has a first and middle name only-no last name. Almost any names for girls or boys can be used. Including a last name could expand the combinations by millions.

In accordance with the invention, repeating the same name for both first and middle names is prevented, e. g. there are no dolls that say,"My name is Susan Susan." Different sets of name elements can be recorded and used in each of the master voice, for example, one set for the Spanish voice family and another set for the English-type of voice family. This leads to even more variation and closer approximation uniqueness across the entire product line.

A culture content editor can select suitable name elements to match to the master voice. For languages other than English, suitable children names for each country may be used.

1. Examples of Girl's Names: Susan, Colleen, Julia, Greta, Violet, Alice, Nancy, Elaine, Stacy, Patsy, Anita, Sharon, Jennifer, Angela, Kathy, Katie, Laurie, Marcy, Lauren, Jessica, Jackie, Christine, Anna, Joan, Barbara, Lillian, Connie, Debbie, Donna, Martha.

2. Example Name Combinations: Hi, my name is Susan Julia. My birthday is June eighth.

Hi, my name is Colleen Greta. My birthday is June seventh.

Hi, my name is Alice Lillian. My birthday is June ninth.

Hi, my name is Anna Susan. My birthday is June ninth.

Favorite Colors Favorite colors can be taken from an artist's palettes. There are thousands to choose from, but the poetic ones are preferable, e. g. amber, turquoise, burgundy, maroon, teal, green, and sky blue.

Each master voice family preferably uses a different set of colors.

Number of Unique Combinations for Each Master Voice For each master voice, different combinations of name elements (first, middle, and/or last names) and/or speech elements may be used to expand the uniqueness or near-uniqueness (apparently-unique) factor over the entire line of product. Also, different speech elements, such as favorite"somethings" (e. g. favorite food, favorite animal, etc.), or lucky"somethings" (e. g. lucky color, lucky day, etc.) may be designed and used.

With just forty name elements and birth dates (month and dates), there are over half a million unique possible combinations for each master voice. If we factor in just two other speech elements such as a favorite color and a

lucky number, the unique possible combinations exceed forty million per master voice, or over three hundred million using just eight master voices.

1. Calculation: Using forty (40) first and 40) middle names = 1560 Using 365 birth dates = x 365 Subtotal of combinations = 569, 400 Using ten lucky numbers = x 10 Using eight favorite colors = x 8 Total unique combinations = 45,552,000 Number of Unique Combinations if there are Eight (8) Master Voices in a Product Line For each master voice family x 8 Total unique combinations = 364,416,000 If a toy manufacturer desires, a talking toy 100 does not have to be unique, meaning that more than one toy can say the same exact speech as another toy. This means that at least two toys may have the same exact composition of speech elements (types, values, and master voice) such as name (first and middle) and birth date.

Considering that the entire population, for example, of the United States is approximately two hundred seventy- five million (not all of them are children and not all of them will purchase a talking toy), an apparently unique or near-unique talking toy will suffice. The manufacturer generally needs to make only a sufficient number of talking toys such that talking toys with the same combination of speech element types or speech do not exist within a certain population, e. g. a local neighborhood or within a state. The end-goal and result is for each child or owner of a talking toy (in accordance with the invention) to feel that he or she has a special toy, which is different in significant ways from all the other toys of that type or product line. By "apparently unique"or"near-unique"we mean that the composition or speech appears to be unique to a typical average end-user. Thus, although the owners of these talking toys, particularly children, view these talking toys as unique, they are in reality only apparently unique.

Considering that children are typical end-users of these toys and children know a limited number of people (friends, family, schoolmates, etc.), a toy manufacturer can feasibly create a number of talking toys all saying the same speech without destroying the apparently-unique quality of these toys. Measures to protect this

apparently-unique quality may also be taken, such as ensuring that toys that have the same speech are never all shipped to the same state, i. e. toys having the same speech are each shipped to a different state. A toy shown in a TV commercial saying a particular speech is only produced once and never sold, or the actual production or marketed toys have a different combination of speech elements from the one shown in the commercial.

Toys saying the same speech are embodied in different looking toys, i. e. each toy looks different (bear, cat, girl, etc.) Considering that a number of speech element types may be combined to form a unique or apparently unique speech (including sounds and music), a product line with only one master voice is feasible. By having an apparently unique toy (rather than unique), duplication may be made and a less stringent quality control may be imposed in the toy-manufacturing process. This may reduce some manufacturing costs.

Serial Numbers (S/Ns) In the preferred embodiment of the invention, serial numbers are used and each talking toy is assigned its own serial number. A serial number herein is not limited to strings of numeric characters. It also includes strings of alphabetic characters, strings of nonalphabetic and

non-numeric characters (e. g. commas, symbols, control characters, caret sign, etc.), and mix strings of alphabetic, numeric, and/or nonalphabetic and non-numeric characters.

An S/N is preferably unique only within the toys of the same master voice family. If a manufacturer wishes to manufacture a product line with eight master voices, the manufacturer may decide to use an S/N, e. g."13422," once in each master voice. Assume further that S/N 13422 corresponds to the speech"Hi ! My name is Sue Ellen and my birthday is October fifth. What's yours? My favorite color is burgundy, and my lucky number is five. Let's play !" ; this speech-i. e. concatenation of speech elements-may only be spoken by one toy in a master voice family. If there are eight master voices, the toy planner may decide to make eight dolls with the same speech, but each toy speaks with a different master voice, such as with a Georgia Peach drawl, Boston Yankee voice, New York voice, etc.

In another embodiment of the invention, S/Ns are unique across all the master voices in that product line.

This, however, decreases the number of unique toys because an S/N is only used once within a product line.

In the preferred embodiment of the invention, an S/N device 114 contains at least twenty-eight (28) bits of S/N information. This is assuming that the toy design

uses the speech elements shown in the table immediately below. The content (serial number) of an S/N device 114 is used to identify the various speech elements that are to be spoken by the particular toy. The S/N generally indicates the words unique or apparently unique to that talking toy. The S/N device may contain a different number of bits if the toy design calls for different speech element types and/or numbers of speech elements.

The table below lists word-combinations count for an exemplary master voice, including the corresponding S/N bits for each speech element type. Speech Element Range Number Number of Type of Serial WordsNumber Bits Elements Used for Serial Number First Name One of forty 40 6 Middle Name One of forty, same as 6 First Name (no duplicates) Birth Month January to December 12 4 Birth Date First to thirty-first 31 5 Lucky Numbers One to ten 10 4 Favorite Colors One of eight 8 3 Subtotal 101 28 Element Not Used for Serial Number Key Phrases Hi... 1 Hello... 1 My Name is... 3 My Birthday is... 3 My Lucky Number is... 4 My Favorite Color is... 4 Let's Play! 2 What's yours? 2 Subtotal 20 TOTAL WORD COUNT 121

The S/N device 114 preferably has one hundred twenty-eight bits, having extra bits available for customizing information. It 114 is also preferably a serial electrically erasable programmable read-only memory (EEPROM). In fact with the same chip, one can customize billions of talking toys, provided that the speech chip memory or voice chip 110 is large enough to hold the entire"dictionary"of possible spoken words or speech elements. Small, low-cost serial EEPROMs are available from several semiconductor vendors, such as Microchip Semiconductor. Corp., Motorola, National Semiconductor, and the like.

The S/N device is preferably organized as sixteen words of eight bits per word as follows: Word Use for What Speech Bits to Use Element (Eight Bits per Word) WO First Name six (6) two (2) bits for CRC (cycle redundancy code) W1 Middle Name six (6) two (2) bits for CRC W2 Birthday Month four (4) four (4) bits for CRC W3 Birth Date five (5) three (3) bits for CRC W4 Lucky Number four (4) four (4) bits for CRC W5 Favorite Color Three (3) five (5) bits for CRC W6 CRC Error Code 1 Eight (8) bits W7 CRC Error Code 2 Eight (8) bits

Words Zero through Five (WO through W5) together generally create an S/N. Considering four bits are used in W2 (birthday month),"0000,""0001,""0010,""0011," and"0100,"for example, may indicate"January," "February,""March,""April,"and"May,"respectively.

The bit combinations"000,""001,""010,""011,""100," "101,""110,"and"111"for W5 (favorite colors), for example, may indicate amber, turquoise, burgundy, maroon, teal, green, sky blue, and lilac, respectively. This is assuming that these colors are the ones for that favorite color set. Generally, a particular bit combination in a word indicates a particular speech element value. Using

the bits contained in the S/N device, a toy of a particular product line is made unique or apparently unique.

Various ways of how to use S/Ns may be implemented.

In another example, assume for simplification purposes that a toy planner wants to use three speech element types to form a unique or an apparently unique speech for toys. These types and their exemplary values are listed below.

Speech Element Type Values Hair Color Black (four (4) values) Brown Blonde Red Eye Color Blue (three (3) values) Green Brown Pet Rabbit (two (2) values) Hampster

If the above three speech element types define the unique or apparently unique combination/composition, there would be a total of twenty-four combinations (hair color x eye color x pet (4 x 3 x 2 = 24)). These twenty- four combinations may be identified by an alphabetic- character S/N. The table below lists how these combinations may be identified. Serial Number Speech (Unique or Apparently Unique Combination) (Hair Color, Eye Color, and Pet) A Black, Blue, Rabbit B Black, Blue, Hampster C Black, Green, Rabbit Red, Brown, Rabbit Z Red, Brown, Hampster

In one embodiment, an S/N device 114 is added or is part of the"speech engine,"which in one embodiment is a finished circuit board (Fig. 3).

In accordance with the invention, when the press-to- talk button is activated 202 (Fig. 2), for example, by pushing the button, power is set on and the S/N for that particular toy is read from the S/N device 114 (Fig. 1) at run time by the speech chip 204 (Fig. 2). The speech chip 108 (Fig. 1), using the S/N read, reads the voice chip and assembles the appropriate speech elements (name elements, birth dates, etc.) 206 from the"dictionary"of recorded words stored in the voice chip 110. The voice chip in one embodiment is the mask program ROM of the speech chip 108. This enables the talking toy to say the speech unique or apparently unique to that toy in the

appropriate master voice. Once the speech is assembled, the toy preferably speaks 208 through an electrical audio speaker. This speaker may be part of or externally connected to the circuit board. Other transducers such as earphones or other electronic-to-acoustical wave- conversion mechanisms may also be used.

In one embodiment of the invention, cyclic redundancy check (CRC) error-detection bit codes are provided, in addition to the twenty-eight serial number bits. The CRC error code ensures that the S/N data are read correctly.

Any other method to set the twenty-eight bits of serial numbers could also be used such as cutting twenty- eight wire jumpers (metal wires or traces) on the PCB 112 (Fig. 1). The cut wires or traces indicate open circuits (bit"1") and the uncut ones indicate closed circuits (bit"0"). Each wire or trace represents one data bit in a multi-bit serial number.

For the production process, it is preferable that the S/N device (serial EEPROM) be inserted into a socket on the PCB 112, as this is the most cost-effective method at this time. The S/N devices 114 are programmed by one or more manufacturing master systems (or master processing systems) 500 (Fig. 5). A master system typically includes a standard personal computer running a custom software created to manage the process as further

described below. There is also a hardware unit in the form of an EEPROM programmer for the S/N device.

Speech Chip Requirements The main limitation on how many speech elements (e. g. names and other words) a talking toy can speak is size of the speech chip memory (ROM) 110. More words require more memory at increased cost. If more names and words are used, the capacity of the speech memory should generally also be increased.

In actual use, only a very few of the many names and words stored in the speech chip memory 110 are actually spoken by any particular toy in order to achieve the desired results at the lowest manufacturing costs now available.

The speech data rate to be used in the speech chip 108 is one high enough to ensure high-quality speech and preserve the voice-family characteristics and the charm that they project. Preferably, the minimum speech data rate is 6. 4 kbs (kilobits per second), with eight (8) kbs or even twelve (12) kbs being even more desirable.

The speech chip 108 also preferably contains a simple four-bit microcontroller, which reads the S/N device 114 on the PBC 112 (Fig. 1) using a two-wire interface. The speech chip 108 has at least two pins for input/output (I/O) to the S/N device 114, and sufficient

operating speed and controller instruction set to implement the data read function from the serial EEPROM 114 to get the S/N data. The speech chip 108 is preferably complementary metal-oxide semiconductor (CMOS) and operates on voltage ranges of 2.4 through 5.5 volts direct current (Vdc) provided by two or three battery power sources with sufficient operating lifetime.

The speech chip 108 is preferably supplied in DIE form for low cost chip-on-board (COB)-style manufacturing assembly of the PCB.

Many major semiconductor companies-such as Winbond Electronics Corp., Texas Instruments, Sonics, and others-produce such speech chips.

Circuit Diagram In one embodiment, the speech engine 300 (Fig. 3) includes a programmable speech synthesis integrated circuit (IC) 302, i. e. speech chip 108 (Fig. 1), from Winbond Electronics Corp. with part number W523M02. The circuit IC1 302 is mounted to a PCB using COB die bonding. Preferably, only nine bond wires are required.

This speech chip utilizes a voice-synthesis algorithm to generate various voice effects and sounds. It also has some commands and programmable registers to provide user- programmable functions. This Winbond speech chip 400 (Fig. 4) generally consists of a timing generator 402, a

microcontroller (controller) 404 with a few input and output pins, a speech synthesizer 408, a serially accessed ROM 406 containing the operating program 116 and the recorded speech elements 118 (Fig. 1), and a pulse- width modulated (PWM) driver/digital-to-analog converter 410 (Fig. 4). A controller as used herein includes processors, computational chips, ASICs, microcontrollers, CPU, and the like. Inputs 412, such as from the press- to-talk switch, are received by the speech chip. Outputs 414, such as the unique or apparently unique speech, are sent or outputted.

The speech engine 300 (Fig. 3) further includes an S/N device 114 (Fig. 1), which in this embodiment, is a lkbit (kilobit) (1,024 bits) serial EEPROM from Microchip Semiconductor Corp. with part number 24LCOlB. This S/N device IC2,304 has the serial number (twenty-eight bits) programmed into it as part of the manufacturing process.

Other similar types of memory chip may be used. It is preferable that the speech chips and S/N devices be integrated circuits.

Power 305 to the speech engine 300 in this embodiment is supplied by three 1.5 Vdc alkaline button cell batteries connected in series to provide 4.5 Vdc.

This provides sufficient voltage headroom for the 2.4 Vdc minimum operating voltage of the speech chip IC1 302 and S/N device IC2 304.

Switch SW1 306 is a pushbutton switch located in the voice module. This is the so-called"push (press)-to- talk"button. The switch may also be sewn into a so- called"sweet spot"on the toy, i. e. an ideal spot where the child user is e. g. naturally drawn to hold, squeeze, hug or manipulate limbs or other features of the toy.

A standard electronic voice module, available from numerous suppliers, uses a spring-loaded cover that activates the switch contact when the toy is hugged or squeezed, or a sweet spot is otherwise pushed or actuated.

Sound from the speech chip is amplified by NPN silicon transistor Q1 308, which is a generic NPN type with a beta of one hundred or more. Components R2 (resistor) 310 and C2 (capacitor) 312 form a simple one- pole low-pass filter to reduce the hissing sound side effects of electronic speech. Power supply filtering is provided by capacitors Cl 314 and C3 316.

The chip oscillator frequency is set by the value of resistor R1 318. The 1.2 megohm value provides for 1.5 MHz operation. Alternate operation at three MHz is also possible, using a 2.4 megohm value.

The two-wire serial interface to the S/N device IC2 304 is for clock and data signals. An optional pullup resistor R3 320 is shown to prevent spurious auto- programming of the IC2 chip in the presence of electronic

circuit noise. Another option is an output for bidirectional data read/write.

The entire circuit is preferably built on a two-inch square (five-centimeter-square) single-sided printed circuit board. It may be necessary to use one or two jumper wires on the board.

The S/N device IC2 304 preferably requires an eight- pin"mini DIP"socket, which is soldered to the PCB. No ON/OFF switch is needed because the standby current used is very low.

Operating Program in the Speech Chip IC1 The push-to-talk switch 306 activates the ICI RESET pin 322 (Fig. 3). This ensures that even if the circuit 302 gets into a"glitch lockup"state (i. e. steps out of sequence), the circuit is cleared each time the push-to- talk switch 306 is activated.

The operating program 116 (Fig. 1) contained in the ROM 406 (Fig. 4) of the circuit 302 (Fig. 3) initializes the CPU (controller/processor) and then runs a routine to fetch the serial number from the S/N device IC2 304 with the CRC error-detection codes to process. The operating program in the circuit 302 then decodes the bit stream and looks at each word-group code received from the S/N device IC2 304. Based on the word group code received (forming the S/N), the operating program selects the

specific speech elements, and forms a concatenation, for example, the first name, the middle name, the birthday month and date, the lucky number, and the favorite color.

In one embodiment, the operating program also randomly selects the salutation ("Hi"or"Hello"), whether the lucky number or favorite color is spoken first, and finally the random choice of the ending words (Let's play! or What's Yours?).

The Winbond W523M02 speech chip supports up to two hundred fifty-six word groups in single page mode, so it can adequately handle the accessing of the one hundred twenty one key words (speech elements).

Since the S/N device 304 is on the printed circuit board, the operating program could also count how many times the press-to-talk switch has been activated. This could trigger some additional surprise words after playing, say, one hundred times.

Serial Number Generation To create a unique serial number that can be hardware programmed into the S/N device, a master processing system 500 is used. The system 500 uses a custom software application 502, which is typically run on a personal computer 504. The software application generates unique or apparently unique serial numbers that are stored in respective S/N devices or EEPROMs. It also

prevents duplicate serial numbers from being generated.

This software application 502 can be coded in any programming language such as C, Basic, and assembly language. The software application 502 also provides upon request relevant printed or electronic reports, such as a listing of all the names of toys in a product line, combinations of spoken messages (speech elements), all words, phrases, sentences, or speeches to be spoken, and the like.

A serial number is generated by taking into account the various combinations of speech elements and using the bits and/or the value (e. g."A,""B,"etc.) stored in the serial number to indicate the set of speech elements to be spoken by the unique talking toy. Once the serial numbers are generated, the S/N devices are programmed using one or more EEPROM programmers 506. The serial numbers may be programmed into these S/N devices in a number of ways, e. g. S/N reuse after X times, one-time S/N use, sequential, random, pseudo-sequential, systematic, and the like. Serial number reuse after X times means that once an S/N has been programmed into an S/N device, that S/N will not be used again unless X number of S/N devices are programmed with other S/Ns.

One-time S/N use means that once an S/N is used, it is taken out of the pool of available S/Ns to be programmed.

Sequential means that S/N devices are programmed with

S/Ns that are sequential in order (considering the bits and/or value of the S/N). For example, after the S/N"A" is programmed into one or a set of S/N devices, the S/N "B"is then programmed, and so on. Random means that S/N devices are programmed with S/Ns that are random in order.

The software application 502 also controls the EEPROM programmer device 506, and in the case of ganged devices for multiple replications, provides process guidance. Using multiple"gang"EEPROM programmers is preferable to make a maximum number of S/N devices per unit time. The EEPROM programmers are connected to the personal computer 504 using an RS232 serial link 514.

Blank EEPROMs or S/N devices 512 are inserted, either manually or via an automatic loader, for programming into the programmer box. Programming time is estimated to be twenty seconds or less, with total handling time per S/N device estimated to be one minute or less. Production capacity per day per processing system is estimated to be one thousand S/N devices (sixty S/N devices per hour x sixteen hours (two shifts)). Once programmed, an EEPROM is inserted into a socket on a PCB 508. The finished PCB 510, assuming the speech chip is already present, is inserted to a toy.

To enhance production throughput, there can be several processing systems 500 running at the same time,

one system for each master voice family. The systems 500 can be connected via a network, e. g. a LAN, to synchronize them.

Color-coded chip identification labels or ink marking, matched to master voice, may also be used to distinguish chips of one master voice to another.

An operator may start the system by entering a master voice family code for the batch of chips to be programmed on a work shift. The software application may keep audits and keep track of all serial numbers used.

Replication of Serial Numbers In theory, each S/N can be programmed into the number of master voice family desired at one time, e. g. eight, using a gang programmer. Then, each duplicate S/N device is placed into a different master voice printed circuit board. Color-coding can be used to prevent placing the same S/N in two toys of the same voice family. If apparently unique toys are desired, serial numbers may be used more than once in a master voice family.

Assuming that the specification (design) of a product line of talking toys is made, many of the operations to produce this line may be made in parallel (Fig. 6). The toy specification may include the kind of

toy (e. g. doll, car, or puppet), the speech that will be said, and the speech elements that will be unique or apparently unique to each toy.

One leg of the production involves manufacturing the toys that will house the finished printed circuit boards 612. Another leg is the generation of the unique or apparently unique serial numbers 608 that are programmed into memory devices 610, such as EEPROMs (S/N devices).

The serial numbers are generated as previously discussed.

Depending on the number of master voice families identified in the toy specification, several recordings of speech elements may be made by professional talents.

Each recording is digitally recorded and captured 602.

The digitally recorded speech elements of a master voice family and the speech chip operating program preferably are written into memory devices, such as ROMs 604. One set of ROMS, for example, contains the recording spoken with a Georgia Peach drawl and the operating program, another set contains the recording with a New York voice and the operating program, another set with a Boston voice and the operating program, and so on.

The operating program generally enables the speech chip to read and decode the serial number, and to assemble the appropriate speech elements. The ROMS are then integrated into the speech chips 606, i. e. a ROM is

integrated with a speech chip (e. g. from Winbond Electronics Corp.).

One speech chip (IC1) and one S/N device (IC2) are then placed into a PCB to create a finish PCB 614. A finished PCB is placed into each toy 616. The toys are then packaged and placed in cartons 618.

Assuming that a manufacturer desires to have eight master voices (Boston 712, Georgia 714, New York 716, Chicago 718, Minnesota 720, Texas 722, California 724, and Spanish 726) in a product line of dolls with each doll unique (Fig. 7)-for example, no same doll with the same name (first and middle), birth day, and master voice family-eight PCBs are prepared 710. At one stage of the manufacturing process, these PCBs 810,820, 880 (Fig. 8) each contain a speech chip 812,822,882 containing the appropriate recorded speech elements and the operating program recorded in ROM (based on master voice family (MVF)) 814,824,884, but without an S/N device 816,826, 886.

At another stage of the manufacturing process, the assemblers have eight S/N devices. These S/N devices, in this embodiment, contain the same serial number, e. g.

"12345,"programmed into them by a master processing system 770 (Fig. 7). The assemblers then insert an S/N device 816, 826,886 into each PCB 730, in order to have a finished PCB 810,820,880, i. e. containing a speech

chip 812,822,882 and an S/N device 816,826,886 (Fig.

8). The assemblers then place the finished PCBs into the appropriate dolls to create eight unique dolls 740- Boston doll 742, Georgia doll 744, New York doll 746, Chicago doll 748, Minnesota doll 759, Texas doll 752, California doll 754, and Spanish doll 756 (Fig. 7). They are then packaged 760, e. g. in a carton containing these eight unique dolls, for distribution.

Packaging and distribution may also ensure that the toys are at least apparently unique. For example, only X number of toys with a speech element value, e. g.

"Susan,"are shipped to each state; randomly picked toys are shipped to each distribution center; randomly picked toys are packaged in each carton, and the like.

In another embodiment of the invention, a library of master voices that includes one, some, or all of the master voice families in a product line is created and stored in a memory device, preferably ROM. The library or sublibrary may include different master voices and number of master voices (e. g. Georgia and Boston), different speech element types and number of speech element types (e. g. birth day, name element, favorite color, and so on), and/or different values or entries on each speech element (e. g. favorite color for Boston master voice is"blue"and favorite color for New York

master voice is."red"). Each voice family may also have all or some of the speech element types defined.

Assume that the product line has eight master voices (Fig. 9) (Boston 906, Georgia 908, New York 910, Chicago 912, Minnesota 914, Texas 916, California 918, Spanish 920), a finished PCB 900 in this embodiment would include a speech chip 902 and an S/N device 922 (similar to the previously discussed PCBs 810,820,880 (Fig. 8)). The ROM 904 (Fig. 9), however, may include the entire library of master voices (Boston 906, Georgia 908, New York 910, Chicago 912, Minnesota 914, Texas 916, California 918, Spanish 920) or subsets of it (sublibrary). The serial number in the S/N device 922 also includes an additional word, W8, that is used to indicate the particular master voice family used by that talking toy. This way the toy speaks the proper speech elements in the proper master voice.

In this embodiment, if each toy has the same exact speech element information contained in its memory device 904, the serial number information contained in the S/N device 922 enables the toy to be unique or apparently unique in that product line.

In another embodiment of the invention, serial numbers are not necessary. A finished PCB 1000 (Fig. 10) includes a speech chip 1002, which contains or interfaces with a memory device 1004. This memory device is

preferably a programmable type, such as an EEPROM or a one-time-programmable (OTP) memory.

Considering the various combinations of one or more speech elements, the memory device 1004 contains generally only the speech elements unique or apparently unique to that particular talking toy 1006. Other information, such as CRC bits may also be stored.

Because generally only the speech elements particular to that toy are stored in the memory device, a look-up of the dictionary of speech elements using a serial number or any other indexing scheme (direct or indirect) is unnecessary. Even if speech elements not particular to that toy are present in the memory device, the operating program instructing the controller in this speech chip knows (firmware, software, hardware) to use only the speech elements unique or apparently unique to that toy.

S/N devices, however, may be added for some other purpose such as to uniquely identify a talking toy within a product line, for example, for speech elements update, recall notices, and the like.

The speech elements unique or apparently unique to a particular toy may be programmed into a preferably programmable memory device by a system similar to the master processing system (Fig. 5). Instead of serial numbers being generated and programmed into S/N devices, the speech elements unique or apparently unique to the

toys are programmed into respective memory devices. The combinations may also be generated in various manners.

The talking toys may also be assembled in a manner (Fig.

7) similar to how talking toys with S/N devices are assembled.

Core Voice Manipulation In another group of preferred embodiments of the invention, a recording for a master voice family is used as a core voice to generate subfamilies of voices 1102 (Fig. 11). These subfamilies of voices are created through electronic signal manipulation of the core voice recordings 1104, for example, using digital signal processing (DSP). Using the core voice, various acoustic characteristics of the recorded voice, such as frequency, phase, pitch, passband, and the like are altered to obtain new subfamilies of voices 1106. Phonemes may be elongated or truncated at either end, or in the middle ; and frequency filtering added to create these subfamilies, which may then be used as other voice families within a product line. The process in this embodiment is similar to that previously described above, except that some or all of the voice families are manipulated digitally from a single core master or smaller number or core master voices.

Depending on toy design, a product line of different-looking toys may be manufactured to fit the sounds of the voice family. As with the manufacturing stages (Fig. 7) discussed above, the finished circuit boards 1210,1220,1280 (Fig. 12) in this embodiment contains two ICs, i. e. a speech chip 1212, 1222,1282 and an S/N device 1216,1226,1286. Each of these speech chips has a memory device, preferably ROM, containing the operating program and the recorded digitally manipulated speech elements (entire library or sublibrary of master voices in the product line). The S/N devices 1216,1226, 1286 are also read for S/N information to determine the speech elements unique or apparently unique to that talking toy (including master voice type, if necessary).

The speech chip assembles the voice elements to create electrical oscillations corresponding to speech, which is then reconstructed through a speaker.

Considering that some of the digitally manipulated voices may sound unhuman-like, the physical appearance of each toy in the product line may be different. The product line, for example, may have toys that look like human beings, nonhuman or imaginary beings (e. g. space aliens, deep-sea creatures), animals (e. g. dogs, teddy bears, cats), and inanimate objects (e. g. lamps, toasters, books, cars, trucks, and trains). In this example, the S/N"12345"is unique only within a voice

family. As discussed above, the entire library or a sublibrary of master voices may also be incorporated in the memory device.

As previously discussed, serial numbers and SIN devices may be eliminated if the memory device contains generally only the speech elements unique (or apparently unique) to that particular talking toy.

Text-to-Speech Synthesizer In another preferred embodiment of the invention, the speech chip 1300 (Fig. 13) includes a text-to-speech synthesizer 1310, either in firmware, hardware, or software, but preferably in software. The speech chip 1300 is preferably an integrated circuit, which includes a controller 1304 that also functions as a speech and sound-synthesis controller. The controller 1304 may also include a digital-signal processor (DSP) circuit 1312.

The DSP 1312 processes digital data representing sound and speech information which are then outputted to an audio transducer such as a loud speaker, earphone, or other electronic-to-acoustical wave-conversion mechanism.

The speech chip 1300 may also include a timing generator 1302 and a D/A converter 1314. Inputs 1316 and outputs 1318 are also handled by the speech chip 1300. Text-to- speech generation algorithms, including programs, and

digital-signal processing are available and known to those in the art.

The speech chip 1300 of this embodiment also interfaces with an S/N device to select speech elements unique or apparently unique to that particular toy. S/N devices, however, may be eliminated if generally only the unique or apparently unique speech elements are stored in the memory device.

The memory device 1306, in this embodiment of the invention, stores/contains the speech element values as text 1308 (ASCII, EBCDIC, ISO standards, etc.) rather than in digitized sound form. By storing the speech elements as text, the capacity requirement of a memory device may be substantially reduced, particularly when the dictionary of speech elements is quite large. For example, a typical average eight-letter word requires eight bytes of digital-data memory storage if the word is stored in ASCII text format. This same eight-letter word, e. g."tomorrow,"representing approximately half a second of sound, if stored at a sample rate of eight thousand samples per second of eight-bit data bits (one byte) per sample in pulse code modulation (PCM) form, would require four thousand bytes of digital data storage. Even with audio data compression techniques, such as adaptive pulse code modulation (ADPCM), delta modulation, or other methods of data compression, and

sound data reduced by ninety percent (and yet maintaining an acceptable quality level in the reproduced speech), the storage requirement is still four hundred bytes.

This is still many times the eight-byte storage requirement of the same word in text form.

Because of efficient text storage in this embodiment of the invention and given the same amount of memory storage, a talking toy may both say many more words and make even more unique or apparently unique combinations of speech elements, such as more unique or apparently unique sentences. Although the memory required to store a text-to-speech-converter program (if in software) may be significant, it may still be much less that the total memory required to store a number, e. g. ten thousand, of words in digitized sound waveform.

In another embodiment, the text-to-speech-converter program enables various parameters of the synthesized voice, speech, or sound to be electronically altered by control of acoustic characteristics such as pitch or frequency of speech, rate of speech, inflections of speech, and similar acoustical parameters. In this embodiment, serial numbers may also be used to control speech parameters, thereby facilitating production of additional voice variations to further enhance the character or personality of voices, as well as create a number of master voices.

Text-to-speech control programs designed to execute on DSP controllers or on controllers using software algorithms may also be based on real voices, such as voices of males, females, youth, infants, teenagers, adults, elderly humans, and so on, and also on cartoon- like voice talents. Thus, a number of master voices may be used in a product line.

By choosing a vocabulary or dictionary of five thousand or even just five hundred words, the complexity and memory required to store the operating program of the text-to-speech converter may be reduced so as to fit in the relatively small memory storage device of a typical toy product.

The text-to-speech converter is preferably operable within a memory size of roughly sixty-four to one hundred twenty-eight kilobytes, on an eight, sixteen, or thirty- two bit controller (CPU) running at a clock speed of as low as four to about thirty-two MHz.

Text-to-speech software-product vendors, such as Fonix Corp., Sensory Systems, L & H, and Dectalk, are available.

Changing of Voice (Aging and Morphing) In another embodiment of the invention, the voice of the talking toy may be"aged"or changed with time by the incorporation of a clock-calendar, preferably in

software. Another way of time measurement is by measuring and storing the amount of play time with the talking toy, e. g. by using a motion-sensing switch, which registers pulses to the input of the controller, counts the pulses, and stores the accumulated pulse count back into some bits of a non-volatile memory device, e. g. S/N device. Based on such time measurements, some of the speech characteristics may then be varied (even in real time), or a new or revised set of prerecorded speech elements in sound waveforms is used, thereby enabling the voice of the talking toy to change slowly over time- over hours, days, weeks, months or even years, such that the voice appears to change and grow older.

Considering that sound characteristics (acoustic parameters), such as pitch, speed, and inflection, may be changed, the voice may also reflect different moods (happy, sad, sleepy, wide awake, and so forth) and other realistic voice characteristics found in human and animal voices.

The present invention has been described above in terms of a now-preferred embodiment so that an understanding of the invention can be conveyed. There are, however, many configurations for speech-engine devices and unique or apparently unique talking toys, not specifically described herein but to which the present invention is still applicable. The foregoing illustrates

preferred embodiments of the invention by way of example, not by way of limitation.

For example, the speech chip may come from another manufacturer, the recorded voices may be stored in a ROM external to the speech chip but with the speech chip able to read that external ROM, or the set of speech elements for a particular toy is stored in more than one memory device. The S/N devices may be recorded in another type of memory device, such as a flash memory or a one-time- programmable read-only memory, the ICs may be mounted on the PCB in a different way, the packaging (e. g. (DIP) packaged form rather than DIE form) of the chips may be different, a different set and/or number of speech elements may be used, speech elements may be phonemes rather than full words, and so on. A person skilled in this field will recognize that such variations may exist without departing from the principles of the invention.

The present invention should therefore not be seen as limited to the particular embodiments described herein, but rather should be understood to have wide applicability with respect to speech engine devices and unique or apparently unique talking toys. All modifications, variations, or equivalent arrangements and implementations that are within the scope of the attached claims should therefore be considered within the scope of the invention.