JPS6159399A - Voice memory - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to an improvement in a speech memory in which speech waveforms are PCM coded and stored in predetermined units such as words or phrases.

[Technical background of the invention]

For example, a voice response device that uses a telephone line to provide various automatic notifications such as time signals, weather forecasts, reservation status, etc. is generally equipped with a voice memory device, and each voice waveform information stored in the voice memory device is stored in advance. It is configured to transmit a desired voice message by selectively reading out and sequentially reproducing the voice message.

Incidentally, in recent years, as voice memory devices used in voice response devices, there are devices in which voice waveforms are divided into predetermined units such as words and phrases, and each of these voice waveforms is PCM encoded and stored in a memory.

This type of voice memory device stores and reproduces voice waveforms as they are, so it has the characteristic that the reproduced sound quality is extremely superior compared to devices that memorize, for example, rules for voice synthesis. .

[Problems with background technology]

However, conventional voice memory devices of this type generally store only the voice waveform portions that constitute words or phrases, and read these out to reproduce the voice. This tends to cause shortcomings such as the beginning of words being cut off or unclear.

[Purpose of the invention]

The present invention is capable of obtaining high-quality reproduced audio without causing any deterioration in the rise characteristics of the reproduced audio, and which can be achieved extremely easily by only partially improving the existing encoding method. The purpose is to provide a memory device.

[Summary of the invention]

In order to achieve the above object, the present invention provides that the current PCM encoding method, which converts an audio waveform into different code strings in a positive audio signal region and a negative audio signal region, has a predetermined level of an audio signal. Focusing on the point expressed by the PCM code of the street,
A non-speech portion is added to at least the beginning of each word of the speech waveform, and the predetermined level is expressed as the predetermined level of the speech waveform portion of each speech waveform block obtained thereby. By assigning one code to the non-speech portion and assigning the other code to the non-speech portion, each audio waveform block is PCM encoded and stored in the audio memory device main body, and the PCM code string read from the audio memory device main body. By detecting the code assigned to the non-speech portion from the PCM code string, a signal for discriminating between the speech waveform portion and the non-speech portion of the PCM code string is output.

[Embodiments of the Invention] One of the currently used PCM encoding methods for audio waveforms is the μ-LAW method. For example, as shown in Figure 1(a), this method has nonlinear analog-to-digital conversion characteristics, and converts the analog input into different independent PCM code strings for the positive signal region and the negative signal region (positive region is '80''~'FF', negative area is 'o
o ” to “7 F ”), and the analog input OV (predetermined level) has two signs (” F
F'', '7F''). Incidentally, FIG. 1(b) is an enlarged view of the above conversion characteristic near OV. This embodiment focuses on the fact that two types of codes are assigned to the OV of the analog input signal, and adds a non-speech portion to the beginning of each speech waveform divided for each word or phrase. When encoding the audio waveform block configured in this manner using the μ-LAW method, '7F' among the codes representing the OV is assigned to the OV of the audio waveform part of the audio waveform block, and - the voiceless part The audio waveform block is PCM encoded by assigning 'FF''' to 'FF', and this PCM code string is stored in the read-only memory (R
OM).

FIG. 2 shows an example of a circuit block diagram of an audio memory device using a ROM in which each audio waveform block is stored in this way. , R.O.
It is comprised of a parallel/serial converter (P/S) 3 that converts the signal form of the PCM code read from M1 from parallel to serial, and a discrimination circuit 4. As shown in FIG.
Performs M code calling and parallel/serial conversion. On the other hand, the determination circuit 4 determines whether or not the PCM code read from the ROM 1 is 'F F '' representing a non-speech portion.If it is 11 F F T+, it determines that it is a non-speech portion. In other cases, a discrimination signal ES of 0'' indicating that it is a voice waveform portion is output.

Therefore, when such a voice memory device is used to make a voice response, for example, the voice reproduction circuit (decoder) decodes the PCM code according to the level of the discrimination signal ES output from the discrimination circuit 4. By setting the timing and decoding, it is possible to reproduce high-quality audio without causing interruptions in the beginning of words in the audio waveform. for example,
When phrases and words are successively read from the ROM 1 and reproduced to synthesize a voice message as shown in FIG. When this signal is detected, the discrimination signal ES of "H" level is outputted. Then, when this signal arrives, the audio reproduction circuit opens the PCM switch and thereafter outputs the PC from the parallel-to-serial converter 3.
The M code string DS is sequentially decoded to reproduce the audio waveform.

Therefore, for each phrase or word, the unvoiced part added to the beginning of the word (A, C in the figure) is played back, and then the speech waveform part (Beginning, 2 in the figure) is played back. Even if there is a start-up delay in the audio playback circuit, this delay will not cause problems such as the beginning of a word in the audio waveform being cut off or becoming unclear, and it will be possible to reproduce a clear and easy-to-hear audio waveform. , can send out high-quality voice messages. Furthermore, since the non-speech portion is distinguished by an independent code 11 F F +' that is different from the audio waveform portion, even if the audio waveform contains an OV portion as shown in the reproduced audio waveform OM in Figure 4. Even if there is, this OV portion will not be mistakenly recognized as a silent portion. Furthermore, we focused on the fact that the conventional μ-LAW system represents the analog input OV using two codes, and used one of these two existing codes to represent the silent part. Thereby, there is no need to set a completely new code, and as a result, there is an advantage that it can be realized extremely easily without changing the PCM code itself at all.

FIG. 5 shows an example of a configuration in which the voice memory device of this embodiment is installed in a private telephone exchange, and this exchange has a PCM code decoding circuit in each extension circuit 61, 62 and central office line circuit 63. The decoding circuit decodes the PCM code string output from the audio memory device M10 and reproduces the audio waveform. For example, when making a wake-up call to the extension telephone 20, the control circuit 64 first issues a connection control instruction to the extension circuit 61 corresponding to the extension telephone f1120 to form a communication path and make a call. When the extension telephone 20 responds to the call, the switching circuit 65 is controlled to connect the extension telephone 20 and the voice memory device 10. Then, the control circuit 64 outputs a control signal indicating the initial address value etc. to the audio memory device ft10 to read out desired audio blocks sequentially. This audio block is then led to the extension circuit 61 via the exchange circuit 65,
The signal is decoded by a decoder provided in this circuit and reproduced into a voice waveform, which is then sent to the handset of the extension telephone 20 and broadcast as a voice message.

At this time, a signal indicating a non-speech portion is output from the voice memory device 10 to the control circuit 64, and the control circuit 64 converts the PCM code of the voice waveform portion outputted from the voice memory device 10 after this signal to the exchange circuit 65. via extension 11
20 to control the communication path formation. Therefore, the handset of the extension telephone 20 outputs a high-quality voice message without interruptions at the beginning of words. By the way, according to this embodiment, since the extension circuits 61 and 62 of the private branch exchange are equipped with decoders in advance, the voice waveform can be reproduced without adding a new decoder, which is economically advantageous. effective.

Note that the present invention is not limited to the above embodiments. For example, in the above embodiment, of the two codes representing the OV of the analog input, 'I FF 11 is assigned to the silent part, but LL 7 F T+ may be assigned. In addition, in the above embodiment, a silent part of the length to be actually output is added to the beginning of a word or phrase and stored. A sufficient length of a silent portion may be added, and the length of the silent portion to be actually reproduced and outputted may be set and varied as appropriate by the audio reproduction circuit. In this way, the intervals between each word or phrase can be set optimally, and a voice message that is easier to hear and clearer can be obtained. In addition, the configuration of the device, the field of application, the method of storing data in the voice memory device body, etc. can be modified in various ways without departing from the gist of the present invention.

〔Effect of the invention〕

As detailed above, the present invention provides that the current PCM encoding method, which converts an audio waveform into different code strings in a positive audio signal region and a negative audio signal region, converts a predetermined level of an audio signal into two types. Focusing on the point expressed by the PCM code, a non-speech portion is added to at least the beginning of each speech waveform divided into predetermined units, and a predetermined speech waveform portion of each speech waveform block obtained thereby is added. assigning one of the two PCM codes representing the predetermined level to the level and assigning the other code to the non-speech portion to PCM encode each audio waveform block and store it in the audio memory device main body; And by detecting the code assigned to the non-speech part from the PCM code string read from the main body of the audio memory device, a signal for discriminating between the speech waveform part and the non-speech part of the PCM code string is outputted. This is what I did.

Therefore, according to the present invention, it is possible to obtain high-quality reproduced audio without causing any deterioration in the rise characteristics of the reproduced audio, and this can be achieved extremely easily by only partially improving the existing encoding method. An audio memory device that can perform the following steps can be provided.

[Brief explanation of drawings]

FIGS. 1(a) and 5(b) are diagrams showing the conversion characteristics of the μ-LAW method as one of the PCM encoding methods, and FIGS. 2 to 5
The figures are for explaining the voice memory device according to the embodiments of the present invention, and FIG. 2 is a circuit block diagram of the device, and FIGS. 3 and 4 are timing diagrams and signal diagrams used to explain the operation, respectively. The waveform diagram and FIG. 5 are circuit block diagrams showing the configuration when the voice memory device shown in FIG. 2 is applied to a private branch exchange. 1...ROM, 2...address counter, 3...
Parallel-serial converter, 4... Discrimination circuit, 10... Voice memory device, 20... Extension telephone, 61.62... Extension circuit, 63... Office line circuit, 64... Control circuit , 65
...exchange circuit. Applicant's agent Patent attorney Takehiko Suzue (L)

Claims (1)

[Claims] A PCM encoding method is used in which the audio waveform is converted into different PCM code strings in the positive audio signal region and the negative audio signal region, and a predetermined level of the audio signal is represented by two types of PCM codes,
In a voice memory device that encodes and stores voice waveforms, voice waveform blocks are formed by adding unvoiced parts to the beginnings of words, and the predetermined level is expressed at a predetermined level of the voice waveform portion of these voice waveform blocks. Two types of PC
A voice memory device main body in which each voice waveform block is PCM encoded and stored by assigning one code of the M codes and the other code to the non-voice portion, and a PCM read from the voice memory device body. An audio memory device comprising: a code discrimination circuit that detects a code assigned to a non-speech portion from a code string and outputs a signal for discriminating between an audio waveform portion and a non-speech portion of the PCM code string.