使用 WebSocket-Sharp 和 IBM Watson TTS 时出现“接收二进制数据之前需要 JSON 标头消息”错误

Posted 2023-02-16

【中文标题】使用 WebSocket-Sharp 和 IBM Watson TTS 时出现“接收二进制数据之前需要 JSON 标头消息”错误

【英文标题】："Expected a JSON header message before receiving binary data" error using WebSocket-Sharp and IBM Watson TTS

【发布时间】：2019-07-14 18:58:49

【问题描述】：

我正在尝试使用 IBM Watson 的 WebSocket 接口，以便获取音频的单词计时。 Here is a link to the documentation. 身份验证部分似乎工作正常。但我不断收到错误“在接收二进制数据之前需要 JSON 标头消息”。在 OnOpen 方法中发送我的 JSON 消息后。我从 NuGet 获得了 WebSocket-Sharp 库（称为 websocket-sharp-core）。

我该如何解决这个问题？

using WebSocketSharp;
using Newtonsoft.Json.Linq;
using System.Net.Http;
using System.Net;
using System.Net.Http.Headers;

async void GenVoice(string text, string dest)

    Console.WriteLine("obtaining access token...");
    var dict = new Dictionary<string, string>();
    dict.Add("grant_type", "urn:ibm:params:oauth:grant-type:apikey");
    dict.Add("apikey", "<my_api_key>");
    var authClient = new HttpClient();
    authClient.DefaultRequestHeaders.Accept.Add(new MediaTypeWithQualityHeaderValue("application/json"));
    var response = await authClient.PostAsync("https://iam.cloud.ibm.com/identity/token", new FormUrlEncodedContent(dict));
    var auth = JObject.Parse(await response.Content.ReadAsStringAsync());
    var accessToken = auth["access_token"].ToString();
    Console.WriteLine("access token is " + accessToken);

    var client = new WebSocket($"wss://stream.watsonplatform.net/text-to-speech/api/v1/synthesize?access_token=accessToken&voice=en-US_MichaelV3Voice");
    client.Connect();
    client.OnOpen += OnOpen(client, text);
    client.OnMessage += OnMessage;
    client.OnError += (sender, e) =>
    
        Console.WriteLine(e.Message);
    ;
    client.OnClose += (sender, e) =>
    
        Console.WriteLine($"closed; code=e.Code; reason=e.Reason; wasclean=e.WasClean");
    ;


System.EventHandler OnOpen(object sender, string text)

    var client = (WebSocket)sender;

    var message = new JObject();
    message.Add("content-type", JToken.FromObject("application/json"));
    message.Add("accept", JToken.FromObject("*/*"));
    message.Add("text", JToken.FromObject(text));
    message.Add("timings", JToken.FromObject(new string[]  "words" ));
    client.Send(Encoding.UTF8.GetBytes(message.ToString()));
    Console.WriteLine("successfully opened socket");
    return null;


void OnMessage(object sender, MessageEventArgs e)

    if (e.IsText)
    
        Console.WriteLine("got a message of type string, it says: " + e.Data);
    
    if (e.IsBinary)
    
        Console.WriteLine("got a message of type binary");
    
    if (e.IsPing)  Console.WriteLine("got a message of type ping"); 

【参考方案1】：

我找到了解决方案，我认为问题出在 WebSocketSharp 库本身。我切换到 System.Net.WebSockets dll，它现在可以工作了。对于任何需要帮助的人，这是最终的工作代码。

using System.Net.WebSockets;
using Newtonsoft.Json.Linq;

var client = new ClientWebSocket();
var canc = new System.Threading.CancellationToken();
await client.ConnectAsync(new Uri($"wss://stream.watsonplatform.net/text-to-speech/api/v1/synthesize?access_token=accessToken&voice=en-US_MichaelV3Voice"), canc);
var message = new JObject();
message.Add("accept", JToken.FromObject("audio/wav"));
message.Add("text", JToken.FromObject(text));
message.Add("timings", JToken.FromObject(new string[]  "words" ));
await client.SendAsync(new ArraySegment<byte>(Encoding.UTF8.GetBytes(message.ToString())), WebSocketMessageType.Text, true, canc);

var toReturn = new List<Timing>();
while (client.State == WebSocketState.Open)

    var buffer = new byte[4096 * 20];
    var response = await client.ReceiveAsync(new ArraySegment<byte>(buffer), canc);
    var data = new List<byte>();
    if (response.MessageType == WebSocketMessageType.Text)
    
        Console.WriteLine("got a string, it says: " + Encoding.UTF8.GetString(data.ToArray()));