|
|
Part: 65-0098
Category:
Communication
-> Telephony
-> Speech
Description: Speech Recognition Microcontroller
Company: Sensory, Inc.
Datasheet: Download 65-0098 datasheet File size : 362 kB
Request For quote: Find where to buy 65-0098
Datasheet text preview:
RSC-300/364
Speech Recognition Microcontroller
DATA SHEET
General Description
The RSC-300/364, from the Interactive Speech family of products, is designed specifically for speech applications in consumer electronic products.
TM
Features
High-Performance Processor 4-MIPS performance at 14.32 MHz 16 general purpose I/O lines Interrupts, timers and counters Fully static operation; clock rate: DC to 14.32 MHz Highly-Integrated Single-Chip Solution Internal 64 kB of ROM (364 only), 2.5 kB of RAM 12 bit A/D (Analog to Digital) converter Microphone Pre-amplifier Internal 32kHz secondary timer 24 x 24 Multiplier Can store 6 Speaker Dependent words onchip Low Power Requirements Requires single 2.85V to 5.25V power supply ~10mA operating current at 3V Low Power 32kHz oscillator Power-down current less than 5 µA High-Quality Recognition and Synthesis Recognition accuracy: better than 97% (Speaker Independent) and 99% (Speaker Dependent). Synthesis data rates from 5,000-15,000 bits per second 4-voice music synthesis capabilities AGC control compensates for variations in input signal Easily Expanded to Larger-Scale Systems Separate 16-bit Address and 8-bit Data buses compatible with common memory components Separate Code and Data address spaces and memory strobes
The RSC-300/364 combines an 8-bit processor with neural-net algorithms to provide high-quality speaker-independent speech recognition, speakerdependent speech recognition, and speaker verification. The chip also supports speech synthesis, voice record/playback, 4-voice music synthesis, and system control. This CMOS device includes on-chip RAM, ROM (RSC-364 only), 16 general-purpose I/O lines, A/D and D/A converters, a microphone pre-amplifier, and a 4-MIPS dedicated processor. The RSC-300 is designed for ROM-less applications that need more ROM space and consequently use off-chip memory. In addition to providing the horsepower needed to perform speech recognition and speech synthesis, the processor has sufficient cycles available for general-purpose product control. The RSC-300/364 Development Kit allows developers to create custom applications. The Development Kit includes an assembler, linker, simulator, hardware development platform, and library of Sensory technology object code. The highly integrated nature of this chip reduces external parts count. A complete system may be built with only a few passive components in addition to a battery, speaker, and microphone. Low power requirements and low-voltage operation make the RSC-300/364 an ideal solution for battery-powered and hand-held devices. The RSC-300/364 uses a pre-trained neural network to perform speaker-independent speech recognition, while high-quality speech synthesis is achieved using a time-domain compression scheme that improves on conventional ADPCM. Four-voice music synthesis allows multiple, simultaneous instruments for harmonizing. Automatic Gain Control can compensate for people not optimally positioned with respect to the microphone or for people who speak too softly or loudly.
© 2001 Sensory Inc.
P/N 80-0165-O
1
RSC-300/364
Data Sheet
Table of Contents
General Description....... 1 Features ......... 1 Table of Contents .......... 2 Introduction..... 3 Rsc-300/364 Hardware Specifications ........... 5 Using the RSC-300/364 ......... 6 Memory Organization .... 7 Memory Map... 8 General Purpose I/O ..... 9 Interrupts ........ 9 Reset and Clocks ........ 10 Timers and Counters............ 11 Power Down and Wake-Up Operation ......... 11 Analog Outputs............ 12 Hardware Debug Features ... 12 Design Considerations ......... 13 Omni-Directional Microphone.............. 14 Power Consumption and Power Supply Considerations ..... 16 Die Bond Pad and QFP Pin Descriptions..... 17 Die Pad Ring ...... 18 RSC-300/364 Die Bonding Pad Locations ... 19 Absolute Maximum Ratings.. 20 D.C. Characteristics .... 20 Vdd vs. Idd.... 20 A.C. Characteristics (External Memory Accesses) ..... 21 Timing Diagrams ......... 21 RSC-300/364 Instruction Set...... 22 RSC-300/364 Special Function Register (SFR) Summary .. 24 Quality and Reliability........... 38 Packaging..... 40 Ordering Information ............ 41 The Interactive SpeechTM Product Line ....... 42
2
P/N 80-0165-O
© 2002 Sensory Inc.
Data Sheet
RSC-300/364
Introduction
The RSC-300/364 is the newest member in a family of high-performance 8-bit microprocessors featuring a high level of integration, targeted to high-accuracy, low-cost speech recognition applications. The RSC-300/364 is designed to bring accuracy, fast response time and versatility to low-cost, power-sensitive consumer applications. A design goal of the RSC-300/364 was to reduce total system cost while increasing system performance. By including microphone signal amplification, data conversion, recognition and synthesis functionality, and ROM storage (RSC-364 only) with a CPU core on a single chip, dramatic cost and power reductions are achieved. Thus, the RSC-300/364 is able to provide 4 MIPS of integer performance at 14.32 MHz. This allows customer applications to achieve maximum performance at minimum cost. The CPU core embedded in the RSC-300/364 is an 8-bit, variable-length-instruction, microprocessor. The instruction set is most similar to the 8051 family of microcontrollers. The RSC-300/364 processor avoids the limitations of dedicated registers by having completely symmetrical source and destinations for all instructions. Of the 2.5 Kbytes of internal SRAM, 2 Kbytes are organized as a Data Space, and 0.5 Kbytes is for register space. All arithmetic operation instructions may be applied to any register. Any pair of adjacent registers (at an even address) may be used as the 16-bit pointer to either the source or destination for a data movement instruction. Instruction classes allow the pointer to access internal or external Code Space, internal Register Space, or external Data Space. Architecturally, the RSC-300/364's separate data and address buses allow use of standard EPROMs, ROMs, and SRAMs with little or no additional decoding. Provision for separate read and write signals for each external memory space further simplifies interfacing. Creating applications using the RSC-300/364 requires the development of electronic circuitry, software code, and speech/music data files ("linguistics"). This document provides detailed information on those aspects of the RSC-300/364 architecture that are important to product designers and programmers. It describes the physical interface to the chip, printed circuit board layout and other design considerations, the RSC-300/364's instruction set, and memory organization. Refer to the RSC-300/364 Development Kit Manual for information on using Sensory's technology code for speech recognition, speaker verification, speech synthesis, and voice record and playback. Description of vocabulary development ("linguistics") information is beyond the scope of this document and is covered in a Design Note.
Custom Mask Capabilities of the RSC-364
The RSC-300 provides significant and flexible expansion capabilities through the use of external RAM or ROM Products using the custom-mask version of the chip, the RSC-364, may save considerable per-unit cost by avoiding the need for other active devices. The RSC-300 requires an external Code Space ROM memory to contain the program instructions, synthesis data, and Speaker Independent recognition weights. The custommasked RSC-364 with no additional external memory devices must rely on the fixed internal memory for all of its ROM and RAM requirements. The internal ROM in the RSC-364 is application specific, with the amount available for user applications decreasing as the number of synthesis words or other technology usage increases. These finite resources restrict the capabilities of products based on the RSC-364. The product specification for the RSC-364 must be carefully crafted in consultation with Sensory to maximize the use of on-chip memory. Each application will have its own specific limitations, but the table below summarizes some useful guidelines for planning purposes. Not all of the maximums can be achieved in a single custom-masked RSC-364 design. For example, a recognition vocabulary of 40 words may limit the speech synthesis to substantially less than 25 seconds. Note: The RSC-364 (Custom Mask) column assumes no external memory.
© 2002 Sensory Inc.
P/N 80-0165-O
3
RSC-300/364
Data Sheet
Description
Capabilities: Speaker independent (SI) recognition Speaker dependent (SD) recognition Speech synthesis and special sound effects Speaker verification Four-voice music generation Voice record and playback SI Recognition Capacity : Maximum number of words per recognition set1 Total recognition vocabulary size in words, all sets SD Recognition Capacity : Maximum number of words per recognition set1 Total recognition vocabulary size in words, all sets Speaker Verification Capacity : Number of speakers identified per set1 Synthesized Speech Capacity: Maximum total length of all messages Music Synthesis Capacity Number of simultaneous independent musical voices Number of musical octaves available Number of musical tunes available Requirement for custom ROM masks: Custom-masked parts (RSC-364) are not stocked by Sensory
RSC-300
RSC-364 (Custom Mask )1
Limited
Limited support Not supported6
15 Unlimited
15 40 words3
642 Unlimited
63/644 63/5124
642
13/644
Unlimited
25 seconds3
4 2-45 Unlimited
4 2 6
No Internal ROM
Custom masked ROM required
1. 2. 3. 4. 5. 6.
Software for the RSC-364 (Custom Masked) applications may be completely developed and verified using the RSC-300/364 Development Kit and an external 64K ROM memory before committing to an RSC-364 custom ROM mask. Practical limitations to maintain accuracy above 95%. Assumes the use of on-chip ROM/RAM only Assumes external serial EEPROM memory. Depends on choice of musical instrument. Requires external storage for recordings.
4
P/N 80-0165-O
© 2002 Sensory Inc.
Data Sheet
RSC-300/364
Rsc-300/364 Hardware Specifications
Architectural Overview of the RSC-300/364
The RSC-300/364 is a highly integrated device that combines: · · An 8-bit RISC microprocessor. On-chip ROM (64 Kbytes, RSC-364 only), Register RAM (448 bytes), Data RAM (2 Kbytes) and the ability to address off-chip RAM or ROM. Analog-to-digital converter, digital-toanalog converter, and a pulse width modulator. A microphone pre-amplifier
AiFE1 AiFE2 AiNŲ AiN1 AOFE1
A[15:0] EXTERNAL MEMORY INTERFACE SPEECH PROCESSING UNIT D[7:0] -RDC -W RC -RDD -W RD
PRE-AMP
AOFE2 AOFE3
ADC
DACOUT
· ·
DAC ANALOG CONTROL PULSE W IDTH MODULATOR INTERRUPT LOGIC OSC1 TIMER1 TIMER2
2K TECHNOLOGY SRAM REGISTER SPACE
448 bytes
BUFOUT/ PW M
STACK SPACE
8 levels
XI1, XO1
The RSC-300/364 has an external memory interface for accessing external RAMs, ROMs or other parallel memory devices. The RSC364 also has an internal ROM that can be enabled or disabled (partially or fully) by pin inputs (signals -XMH, -XML; See figure 4). ). When the internal ROM of the RSC-364 is disabled, its performance is identical to the RSC-300. With the RSC-364, the entire program must reside in the internal masked ROM. External memory can only be used to store data.
CPU INTERNAL ROM (RSC-364) 32K x 8 32K x 8
HIGH
-XMH -XML
XI2, XO2 OSC2
LOW
PORT 0
TIMING AND CONTROL
-RESET -TE1/ PW M
P0.0-P0.7
BREAK POINT REGISTER
The 8-bit processor can directly access 448 P1.0-P1.7 on-chip general-purpose registers (RAM), and 32 additional Special Functions Registers (SFRs). The instruction set accessing these Figure 1 RSC-300/364 Block Diagram registers is completely symmetrical, allowing movs, arithmetic, and logical operations with any register as the destination. Two bi-directional ports provide 16 general-purpose I/O pins to communicate with external devices (See page 9). The RSC-300/364 has a high frequency (14.32 MHz) oscillator as well as a low frequency (32,768 Hz) oscillator. The processor clock can be selected from either source, with a selectable divider value. Sensory's technology code requires the use of the 14.32 MHz clock. There are two programmable 8-bit counters / timers, one derived from each oscillator. A variety of wait state configurations allow fast code execution and easy interfacing to slow peripheral memories. An inexpensive electret microphone connects directly to the microphone input of the RSC-300/364. The internal preamplifier converts the tiny microphone signal to a level suitable for Analog-to-Digital Conversion. (ADC), The RSC-300/364 uses a Sample and Hold (SH) circuit and ADC converter to convert the amplified analog speech signal into digital data. The chip may also be used with line-level inputs. The output audio signal of the RSC300/364 is derived either from a DAC (Digital-to-Analog Converter) or a PWM (Pulse Width Modulator). In addition to its on-chip ROM (RSC-364 only) and RAM, the RSC-300/364 has 8 data lines (D[7:0]) and 16 address lines (A[15:0]), along with associated control signals (-RDC, -RDD, -WRC, -WRD, -XML, -XMH) for interfacing to external memory. The memory control signals on the RSC-300/364 and the processor instruction set provide independent Code and Data spaces, allowing configuration of systems up to 192 Kbytes with no additional hardware decoding. The RSC-300/364 features 16 general-purpose I/O pins (Px.y) for product and memory bank control.
PORT 1
© 2002 Sensory Inc.
P/N 80-0165-O
5
Others parts begin by 65
65-1
|
|
|
